Ask HN: Software architects – what’s your typical day look like? A question for anyone out there working as a software architect (especially microservice architectures). What’s your typical day comprise of? Where do you find yourself focusing your time? What resources have you found to be most useful? |
Most of my time is meeting. Either other teams or my own.
Between these, I'm either building proofs or managing services we never found owners for with a reorg
Good architects write code or become more disconnected and more similar to management.
I might spend several days (or more) going through code and interviewing developers about how they implemented an authentication protocol so that I can draw it in BAN logic on a slide, which other projects can use to integrate their code with, and our certification/accreditation/infosec people can reason about so the project doesn't get spike stripped at the production gate and delayed a quarter while a post-hoc security analysis gets done on it.
I would meet with our vendors and tech service providers and establish whether they in fact provide the services they said they would, and what that physically means. A great example is whether an o365 tenant supports OAuth2 federated logins with non MSFT tenants or not, and how we are going to enroll users based on the amount of friction the realization of this feature causes(or not). Another one would be pushing standards down into development so nobody would roll their own protocols, and so I could have a short answer to what we implemented. You might see it as ticking a box, but an enterprise is like an airport, and some people get to fly while others don't. The ones who don't are usually stewing over some counterfactual.
I would meet with external consultants, often security and regulatory people, and provide them with the technical details and assurances that would keep them from taking a pound of flesh from the project in the form of a 6 week analysis engagement.
What makes an architect something different is they recognize that when there is direction or demand for a tech project, that creates opportunity for a lot of other very sophisticated technical people to inject themselves into the critical path of the project and use their leverage to extract money, management authority, and other concessions from it. As an architect, you see these people coming, and make sure they do not derail your tech.
I enjoy it because it's solving problems at a higher level of abstraction. I also like doing product, but the architecture urge sabotages that, as in product, solutioning comes at the cost of listening and scaling your listening out to architects to solution stuff.
20% mentoring
60% facilitating consensus between teams/orgs
Then it varies.
- 1:1s
- project or program (in the sense of both code type programs and long term direction programs) planning or status meetings with various stakeholders, especially POs, PMs, and execs
- digging into particularly difficult or odd code or set up issue that's lead to it being escalated
- meetings with vendors and doing a write up of their product's potential and costs
- creating rough project or program outlines with rough budget estimates for execs
- creating design documents on projects or programs that have a bit of traction
- meeting with other architects to collaborate on wider designs or just share items of interest
- going on HN to tell strangers about my day-to-day life
- create fully functional POCs or reference implementations
- doing post-mortems
- examining code from a potential acquisition or quizzing their engineers
- getting on meetings with clients or potential clients to give them a warm fuzzy that a tech person with an impressive sounding title is listening to show we're taking something Very Sersiously and find them to be Very Important.
- regular Open Office Hours for anyone in the company to drop by and chat about anything (lots of good stuff comes from these -- highly recommend)
- fielding various requests and questions from Security and Ops other teams and whatnot.
- Having What-If meetings with sales or execs (danger! There be dragons here! But, with Sales especially, a really good way to figure out what customers are clamoring for)
- Preparing talks and technical outreach
- Exploring New Shiny Things (though I tend to do this more on off-hours just because there are fewer interruptions and urgencies).
In terms of microservices in particular, I tend to ask a lot of questions about data and work flow, how someone has defined a domain or domains, making sure they're not falling into the myriad anti-patterns, making sure what they're doing is going to be visible, tractable, and play nice with the rest of our services including following our common patterns and idioms (and when I run into idioms that they might not know about I document them and make them easy to find) and not create any unpleasant surprises down the road, and that they have given appropriate thought to evolution of the service and forward and backward compatibility during that process.
Martin Fowler is by far my most influential resource on microservice architectures and doing domain driven design well. But I would strongly suggest literally reading everything you can find on the subject. Mostly because it's many subjects in a trenchcoat. Seeing it all from a few different angles will better arm you for whatever comes your way.
On Monday, Wednesday, and Friday I have large no meeting blocks in my calendar to discourage recurring or pre-planned meetings. Most of this time is spent:
1. Writing code
2. Reviewing code
3. Responding to feedback in our Slack channels
4. Impromptu meetings to help other Engineers get "unstuck"
1. In 1:1s with other engineers or team leads
2. In miscellaneous recurring meetings like our "Architects Chat" or "Leadership Meetings"
Reviewing code is a close second. I review roughly 50 PRs every week which is about 2x what our closest "non-architect" engineers review (usually Engineering Managers). Many of my impromptu meetings and Slack feedback comes from what I see in code reviews. If I leave anything beyond a trivial comment, I like to have time to discuss with the individual to make sure I'm understanding correctly and actually providing useful feedback. Since other people are working in my code-bases, its also important to see that "the ideas are coming together". If PRs into our Circuits code are frequently breaking patterns, its likely a problem with my patterns rather than other engineers.
I don't have documentation on the list as it isn't something I write in an ongoing way. Instead, I prefer to write documentation at the beginning and end of large projects. I don't make a ton of edits as I go because the code is in flux so it can be a waste of time to update both code and documentation. Instead I like to:
1. Start a project with a plan of how everything fits together
a. Planning takes anywhere from 1 week to 2 months depending on the size of the project
2. Get approval on the plan
3. Write code... if it has to contradict the plan, so be it
4. Update the documentation at the very end so other engineers can use it as a nice reference for the current "state of the world"
On a typical day, I have 4-6 meetings with different groups of people. Some will be solution design discussions with engineering teams where we try to figure out how a particular challenge will be solved together. Others will be with product managers, talking about feasibility, cost and time estimation. Yet others will be with senior directors about long term strategy and infrastructure. Finally, a very important part of the meetings is mentoring. Knowing something is valuable, sharing that knowledge is invaluable.
While this may sound like a lot of meetings (and engineers typically abhor meetings) they are typically very useful and very rewarding.
The remaining time I typically spend on doing preliminary research, design, documentation and every now and then even coding, which I thoroughly enjoy.
How do you deal with the burden of knowing/predicting what the right thing (tm) is / will turn out to be? I am constantly agonising over choices I'm about to make, and almost invariably come to regret some of the choices that I, or our team, have made (which include, but are not limited to, React, Gatsby, Flow, Stylus, Enzyme, Jest, SCSS, single-page app architecture, possibly graphql, etc. etc...). Not that I'm an architect; but still. How do you cope?
You chose the latest and greatest web framework? Great! How many developers at your company will know how to use it? How long will it take for them to learn it? How big is the talent pool you can hire from? How much time will it take to migrate your products to it? Who will support it, is it "wild west" open-source or does it come with commercial support? How does it fit with your deployment model? Has it gone through rigorous threat modelling? There are about a billion factors that may affect your choices and finding the right solution will always be a central part of the problem.
To actually answer your question about how I deal with this, this is why I do research. Let's say our product people come up with a brilliant new product idea that we don't have an existing solution for, then I'll spend some time on finding information about the available options and take notes on the benefits and drawbacks. I will try to find people who are experts in the subject and get their opinions.
What I can not do, is to actually go and gain a deep understanding of every single language/platform/framework/library there is, I'm just a human after all.
- Keep your code reasonably clean
- Use strong typing
- Make sure your tests don't test implementation details if possible
- e.g. using more integration like tests written with react-testing-library
- e.g. using visual regression tests
- If you sense something is off about the tech stack make a POC of partially migrating to something else as soon as possible
- If the thing you tried is better come up with a plan to migrate and do it as soon as possible
- If you've followed all the bullet points above then switching to a better solution shouldn't be that difficult
By following these guidelines I've never felt on a project that I've painted myself into a corner with the tech stack, or chosen something very bad that is not reversible.
Here's another way of looking at that problem:
How many of those tools and technologies existed 1, 3, 5, 10 years ago?
Would you trust a six month-old tool on github to form the nucleus of your -- probably mundane, in the good way! -- project? Would you risk it all for the chance to have a slightly different way of rendering CSS or generating HTML?
... I didn't think so. React's been around for a long time now and it's got a lot of momentum and people who know it. So that's a pretty safe bet. Rinse and repeat.
Not all tools are critical and you do have some leeway. But when in doubt, pick something you can easily hire people to help you accomplish; that has been around a while; and that feels established.
Let's say you are a great poker player. When you get dealt a good hand, you play it and end up loosing - does that make you a worse player?
Turns out even the best of all players loose quite frequently. What distinguishes them from others is that they make the best choices with the limited information they have at the time. They later go back and learn from any mistakes they have made and move forward.
Choosing between mysql, sqlite and postgresql may be (do we need row level security? Would mysql multimaster benefit us? Do we expect to deploy many, small instances where the simplicity of sqlite will shine) etc.
There are many decisions to make building a solution - the architecture part has to do with fulfilling the requirements.
The tech-lead/senior developer/team lead part has to do with implenting the architecture - our team knows vue, so we go with that to implement the app. We're invested in ms sql so we prefer that to postgres - but sqlite would be a poor fit.
Realize that if you had chosen another solution, you would have other issues. Not "no issues". Probably not even "fewer issues" or "less severe" issues - just as high a chance
Realize that most technology choices actually have very little impact on the business. If one uses A or B, it does not matter much to do whatever it is your company does. Very few tech choices have the potential to kill the company/unit/product.
Realize that once a choice has been made, worrying about it does not help anyone.
So the goals are not perfection. Not anywhere close. Just something that:
1) lets us move forward now (avoid decision paralysis)
2) not worse than before (if so, just stick with what you got)
3) good _enough_ to deliver the needed business value
4) does not paint ourselves into a corner. Then we can always fix and improve things later
I have invested in some of these technologies. The biggest commonality is that they ALL have their faults and weaknesses. The key is ensuring everyone makes best use of whatever framework/package/tooling that you have invested in via a set of best practices and standardised processes. Over the last 20+ years, i've found you should only switch to something else if there are significant gains to be had. Otherwise it rarely offers more value.
You cannot predict the future. It is impossible know which choice will result in better outcomes. Whatever you choose has probably about equal chance of success or failure. So just make a choice, with a general idea of what you will do when the choice fails. Most choices are not make-or-break, so you should really not be agonizing over most choices.
These are cases when the previously mentioned 'professional negotiator' role gets really important. These are the cases when you need to go to the people involved and try to understand their priorities and their constraints. Once you have the information necessary, most of the times you'll be able to come up with an alternative solution that is acceptable for everyone. Sometimes you can convince others to give way, sometimes you'll need to.
And then there are times when you just can't do anything because, for example, you get overruled by an overzealous CTO. In these cases, as sho_hn also pointed out, the best you can do is to make them understand the implications and brace yourself for damage control, because the fallout is going to be on you anyway :)
My company doesn't have a "Software Architect" job title. Instead the closest thing is "Staff+ Engineer", which does pretty much the same things. However there are some key differences.
First, these aren't "Ivory tower" architects. They still write code (although in practice not much) and stay connected to implementation details, mostly via code and design reviews.
Second, architecture is every dev's responsibility, at different levels. So Staff engineers might oversee the architecture of entire systems spanning a whole org, while senior engineers oversee a single team. A mid level dev might design the "architecture" of a small service integrated in a larger system.
Staff+ engineers usually report to senior leaders and aren't part of a team. They might hold office hours and break disagreements within or between teams.
I'm not one myself but in a conversation a Staff eng. told me this was the first job where no one tells him what to do. Instead he spends his time proactively looking to solve business problem. This might be top down (e.g. take a business problem and try to find how to solve it) or bottom up (e.g. take some new tech or tool and figure out how it can be used to achieve some business goals). Often problems can also be solved without tech at all (e.g. aligning stakeholders).
Lastly at this level engineers are expected to be leaders as well. Mentoring, sponsoring, etc. is pretty much a requirement. They should be force multipliers, making other engineers around them better. They might also scale themselves by producing content, such as tutorials, talks, training, etc.
Not off-topic at all, this is aligned with what I see. In older or 'regulated industry' companies the role and career path of Architect still exists. But more and more I just see senior roles that are flavours of Principal/Staff/Fellow and they still have normal Engineering duties but then in early greenfield stages of projects act as Architects, just not as a full time role.
- Spending a lot of time trying to figure out the business better
- Studying a lot of books, articles, codebases and discussions on software architecture
- Inspire the team and younger devs to avoid complex tools and solutions and stick to the basics
- Balancing everyday and urgent business needs while leading towards a more simple, boring and maintainable system.
- Trying to persuade people to avoid technical debt at all costs
Depends on the size of the firm. Very, very large ones will see you write little code -- generally to the detriment of the company, the architect, and the teams they interact with.
Good architects are hands-on. They write proofs of concept implementations before recommending a tool or technology.
Bad architects dream up nothing but gormless and inscrutable charts and diagrams that, like crap ink on vellum, slowly fade into illegibility in a remote Sharepoint server.
Good architects clarify and explain. Their job is to assist teams in melding disparate technical viewpoints with that of the business -- who themselves often lack focus. A good architect facilitates that: they'll rarely look at lines of code, per se, but more the general trend and direction.
Bad architects dream up complications that serve no one but their own ego. I once worked with an "architect" who spent most of his time mapping out some data, represented in JSON, using eBNF -- a tool used for specifying formal grammar for computer languages. It could've been solved with an example + some data typing in the margins. Don't be like that dude; people struggled to make sense of the trivially simple data we had to store. For larger things use a schema-like structure; for smaller things, trust that people and teams are not stupid.
Good architects are skilled at everything. They've hacked up CI systems in batch or bash files, before the current tooling even existed. They've built all manner of systems, in a range of environments and possibly languages. They're pragmatic, and can spot trouble a mile ahead.
Bad architects envelop themselves in what ever the buzz du jour is: like microservices, which is never an opening gambit for 99.9% of problems you're going to encounter in a greenfield project. They're the ace up the sleeve when all other performance and scalability efforts have failed. Good architects know when to tell a tech lead to lay off the technobabble: chroming your CV is not in the interest of the business as a whole.
Good architects prefer plain, old working tools. That usually involves a RDBMS. Sometimes many; and often times you need to make existing ones talk to other things. Bad architects thing old means bad. XML is not bad; what people did with it was, for example.
But once I suggested they should do some back and forth with the people actually writing code... they told me "I cannot understand code, and it is all right so. I do not need to. I'm the architect, not the carpenter"
Another time I asked the chief architect --we were working in an embedded system-- "I have to program this function, what is my RAM, ROM and CPU budget for it?" His reply shocked me: "I'm architect here, I do not care about that little things, I do only the architecture, look here" and he proceeded to show me lots of boxes connected with lines. Then he added "that here is what is important, the real architecture is here".
Needless to say, those projects failed miserably, with no hope for recovery...
The key role that our architect played that brought immense value to us was being the integrator of teams. Teams had a high level of autonomy when it comes to implementation details which meant there could easily be “reinventions”, incompatibilities, and even narrow minded choices. Our Architect always had the big picture view across teams, almost always sat in our feature grooming sessions with valuable inputs, worked closely with product owners who bring in feature requests that may need integration of multiple services etc.
Above all, he was very active, and approachable to all teams instead of hiding behind barriers and occasionally bringing out UML diagrams. Several times, when we were debating which technical choice to make, held chime in with valuable points about what would be better if we were to integrate with another service, or something like “that team did this because blah, which might be relevant to you”.
My day job is basically to be a "force multiplier" and it boils down to:
- cat herding = meetings, discussions, negotiation, "shoulder to cry on" = way more soft skills then I ever thought I'd need
- Big picture stuff = "town planning" for tech, second order thinking, pulling together cohesive plans/strategies, principles, constraints = way harder to effectively communicate this stuff than my younger stuff would have thought
- rapid altitude changes = dropping from the 10000ft view down to helping a team troubleshoot some production issue, helping a junior dev with a code issue, solving a dispute between devs, hands on evaluation of some new tech, then jumping back up to talk to a leadership team about some new grand strategy, or to planning out a multi year program of work = a ton of skills that my younger self would have never have guessed at, finance, budgets, "business" language, operating models along with keeping my technical skills sharp
In terms of resources:
- Anything around learning to story tell and communicate effectively
- The "97 Things Every Software Architect Should Know" essays
- Your tech skills - write PoC apps, side-projects, try out new tech, learn to quickly grok strengths & weaknesses of tech
- 80% what needs to be done today/what fires need putting out, and 20% planning for the future and seeing how the bleeding edge can legitimately help us (e.g. WASM coming soon...). This is part of the general keeping up with advances in industry
- Mentoring, jumping on pair programming calls to unblock other teammates.
Less coding, more meeting and reviewing, overall clear view of the big picture is a must.
Sadly I've seen "architects" being more focused on drawings, crude prototypes and fluffy ideas. Basically dreaming up ideas and concepts, throwing them over the wall to developers and SREs to pick up. A previous boss of mine had to callback a candidate for a job, telling him, that despite us hiring pretty much anyone, we had no need for someone who thought of software architecture as a desk exercise.
Do you have insight on what the proposal, design review, and first commits look like when architecting a brand new project? Like what do the first 30-60-90 days look like from product idea up to those o&m and sustainment activities?
From my experience, the best architects spent most of their time focusing on these things:
1. Always understand what the current business priorities are.
2. Design the overall system architecture so it covers all of the business needs, and nothing more. Always focus on keeping it as simple as possible. Even when people argue "lets do this some other way, because we might need it in the future".
3. Fight the urge to use "latest and coolest" technology. Don't listen to consultants selling you unnecessary tech. Stick to what works. Even when it's boring. Explain it properly to your team.
4. Maintain proper documentation. Focus on making it as simple as possible, easy to adjust and up to date. Otherwise its value will be close to 0.
5. Use UML diagrams. Think about where the complexity of the given system is, and use only the most relevant ones. For example, if the complexity lies in complex workflows, use sequence diagrams or state diagrams. If the complexity lies in complex datamodel, use ERD diagrams.
[shameless plug] That's one of the reasons I founded https://stacktape.com - AWS-focused "internal developer platform" that allows developers to do this sort of stuff on their own, without involvment of architects, DevOps or others.
“Good architects prefer plain, old working tools.” https://news.ycombinator.com/item?id=33880398
“Fight the urge to use ‘latest and coolest’ technology. (…) Stick to what works. Even when it's boring.” https://news.ycombinator.com/item?id=33880978
"Inspire the team and younger devs to avoid complex tools and solutions and stick to the basics” https://news.ycombinator.com/item?id=33880224
I strongly disagree with this. Fun killers are working late nights to remedy poor planning, or getting paged multiple times every night because people didn't make informed choices. It kills the fun for the employees as well as others they interact with.
A good architect includes someone who understands the needs of the people in terms of growth and satisfaction. They'll push back on schemes that introduce toil and try to align around good bets that contribute to your skill set. That may involve blocking use of technology X which has Y long term consequences. The consequences of that long-term choice need to have counters to keep the people engaged, so the good architect considers how to manage that. They don't always get that right and we have some less well-rounded architects, but we should aspire to doing this.
Oh boy we have fun in the 5w and retrospectives when changing one corner of the app crashes everything else and the legal team now is in deep conversation with insurance to cover claims from customers.
/s
* Experienced devs who have "seen it all"
* Have to deal with the consequences and have more responsibility
I have no formal power in the sense that people are not supposed to do what I say, so I have to be convincing, and influence others behaviors to make the software better.
This implies analysis and compromise, having meetings, mentoring engineers, and quite a bit of coding and research.
My work is less guided that when I was a junior engineer, in the sense that I'm meant to discover opportunities for improvement and give guidelines. This leaves time to pick what I should be doing next, where I'm adding the most value.
This changes from time to time, it may be figuring out a new architecture, fixing a bug, getting involved in product development or handling an incident.
An important part of the job is getting involved in production incidents, this helps grasp where are we struggling and opens the opportunity to guide development to a state where customers are happier and operational costs are lower.
You also start to be more conscious about company finances (how much it costs to keep the lights running) and start thinking in ways to get more from the same resources.
Even if I have no formal power, people tend to pay attention to what I say and suggest, so I need to be careful, because an off-hand comment can have a lot of impact.
Architectural roles live and die by their word. We lead by influence, so be cognizant of perception and careful when you communicate with others in every interaction.
Meetings here and there, lot's of talking to people and lots of reading. Currently I'm working on a digital transformation strategy and I can definitely see myself writing a whole lot in the next few months to gather traction on the implementations that are needed.
I would say that at least for me, becoming a Software Architect has meant that I need to split my time between groundwork and strategising with an added layer of complexity that stems from the fact that I work for government, so there are a number of policies and internal bureaucratic barriers to get used to.
In the morning I look over our github/mail/slack queue to see whether someone needs urgent help.
Then, if there are customer meetings scheduled on that day, I frenetically try to re-create customer's infra in our lab (be it Anthos Cluster, AWS Systems Manager, Azure Arc, or istio). If there are no customer meetings I try to improve our stuff on github. Sometimes, I publish new blueprint, kubernetes operator, or a terraform module, or try to create new CFT resource.
Leave the code review to the experts in each codebase — most changes will be on point and need only a small or large amount of alignment before they can land. (By contrast, very few things need to have the brakes put on them.)
Some changes though will correlate with other problems across the codebase and this is where you should be stepping in to spot future patterns or current anti patterns and providing solutions and directions forward, or at worst, road blocks.
Once you have enough of these under your belt you can start proactively spotting hot spots in the eng org that need focused effort. More mypy typing for a core library. Two libraries that should be one. One library that should be two. Vertical slices of functionality in two products that should be horizontal slices pulled into a separate service (or, my preference, library.)
Processes are important too. You’ll see what people are repeating and or finding hard, and which could benefit from some love. Build infra. A testbed for debugging a process that is otherwise too heavyweight with production data. Teams that don’t talk enough pre-PR. Managers that need help managing consistent poor performers.
I think many people wanting to go into an architect career think it is highbrow design work. In reality, I’ve found you end up doing much more support / boiler room / janitorial work. It’s very satisfying, and very much reminds me of my career sidetrack as a school teacher. Above all, you are there to help people and teams reach their potential and get the most from their jobs.
So sometimes we do some design meetings. I suggest how to better design database, etc. I'm trying not to dictate everything but rather catch mistakes which are obvious from my experience. Avoiding mistake early is very important IMO.
Recent months I designed Kubernetes cluster. We have plenty of small and medium services and environments thrown around in a few dedicated servers. It's a giant mess. I dread when I need to untangle it. So it was unavoidable to redesign devops from the scratch and I decided to go with Kubernetes. I basically tried few approaches, some turned out to be too complex, e.g. I built a complete automation on terraform, shell scripts and flux and ditched it out, because nobody but me would understand it and I don't need that kind of job security. I ended up with a simplest setup possible. Terraform provisions servers (we use hoster with OpenStack), then I manually run kubeadm with prepared configs, then I run few shell scripts to install important stuff with helm, then I install our services with kustomize. I think it was very nice outcome. Simple and approachable for everyone who knows basics of Kubernetes. Not full-fledged GitOps, but I decided that we're not ready for it yet. That was one example of project that I did myself because we didn't have necessary expertise.
Right now I'm adapting our projects to work with Kubernetes, improving builds, adding health checks, writing yams, etc.
I also made a foundation for some important core projects which I couldn't trust others to design. Then people continued my work. It seems to work fine so far, I oversee those projects to keep them in a good condition.
Also I chat a lot with developers who stumble upon hard issues and struggle to resolve those or they can't make some decision.
And I have some vision how our system as a whole should look like in an ideal world. We don't have resource to implement that vision and probably never will, but I consider it a good direction so I'm trying to point important decisions to that direction.
Also I sometimes walk over repositories and fix stuff I don't like. Usually devops stuff, like bad dockerfiles, missing dockerignores, outdated dependencies.
Sometimes I feel like a janitor, LoL.
Then I go to meetings and track down people to clarify requirements and do an architecture diagram or two and write stories from it and get it "approved" by the PO.
I also pick up stories the team doesn't want, since I feel responsible for them not being clear / easy / broken up enough. In the evenings, I study for cloud exams.
Resources I have found useful is really understanding flowcharts, and the difference between flowcharts and high-level architecture diagrams.
Early stage days are estimating + research. Often sitting with users and figuring out system complexity (from a usage and implementation perspective). Most important part of any project/transformation. Can never know enough. This is usually 2-3 hours meetings per day with users/technical groups, 3-4 hours documentation pieces.
Later stage projects are process heavy. 3-4 hours with support, incident managers, networking, testers, devs, SLA stuff, NFRs all that fun stuff. Rest of the day writing.
Overall it can be busy with long presentation prep and discovery/planning work or it can be quiet (something gets delayed) where I just go play with some new tech for a few days!
Most of my time was spent on high level design of complex business and B2B products and making sure that managers, business analysts and programmers do their jobs. Also selecting technologies etc. I still had about 50% of my time left for actual work so I would always pick some particular chunk and create design documents with diagrams given to programmers and would also code some chunk. I could have avoided doing any "productive" stuff but I did not want to loose touch. Being converted to a pure manager was not in my nature.
I'm also the last point of escalation for a handful of dev teams. That's the fun part, getting to work on problems, mentor, and help out. I really enjoy mentoring and watching inexperienced people blossom and start mentoring others and passing on what they've learned. It's very fulfilling.
By the way, I highly recommend his book Staff Engineer if you're interested in delving deeper.
Although I'm not an "architect" by title, things I build do impact many teams as we move towards standardising how to do X.
> - Balancing everyday and urgent business needs while leading towards a more simple, boring and maintainable system.
> - Trying to persuade people to avoid technical debt at all costs
This is literally my job everyday now as a principal. I enjoy it, but it's quite grueling, tbh.
"Fundamentals of Software Architecture" is 95% fluff. So is "Clean Architecture". Most patterns like CQRS, DDD, fully event-driven etc are complete overkill and a sign of a BAD architect unless there is a real cause for it. 90% of the job is NOT overcomplicating.
"Trying to persuade people to avoid technical debt at all costs" - put in a linter and concrete rules (even if a rule is loose, note that in writing) or devs will get upset and think you are singling them out.
If you want to write a script for use once or twice, worrying about coupling and cohesion is probably not worth your time. If you are building a system to last 2-5 years out, it absolutely is worthwhile.
If you have a dead simple CRUD problem with very little business rules/operations then DDD isn’t right either. But if you have a complex set of interactions, state, and policies it’s the right call. (The big red book of ddd has a great table on exactly this)
But there are problems out there that are suitable to those approaches. The devil is in the detail; unfortunately a lot of strategic design goes by the wayside to cargo-cult zeal.
But the same can be said for the other way of thinking. I’ve worked/work with engineers who think any form of abstraction is abhorrent and that they don’t need to care about quality because they’ll delete it all and start again (they’ve not tended to care about the opportunity cost of that to their employer, either). Any database interaction looks like ActiveRecord and all technology choices are immutable.
The key in all things is balance. What is simple vs overwrought is almost entirely down to what the framing of the problem is.
I totally agree with this, but
> Most patterns like CQRS, DDD, fully event-driven etc are complete overkill and a sign of a BAD architect unless there is a real cause for it.
Once your application gets to a certain size, DDD, CQRS, and events will make your life far simpler.
Avoiding these patterns is a bit like saying that the solution to moving a piano is to get a bigger bicycle.
Technical debt does not increase only for technological reasons but also from business expectations. You can't just "build" a new garden and leave it be. You have to maintain it also. Keeping the system clean and well oiled before building more is important. But these concepts need to be reminded everyday to survive
I went from a company with no linter or style guidelines and PRs were an absolute nightmare. Eight devs, eight completely different styles, each insisting theirs was right. I wasn't able to turn stones into loaves (i.e. I wasn't a Senior Dev at the time), so my style was categorically incorrect. I switched to a company with an aggressive linter, and all that superfluous debate over tiny patterns and "personal preference" went out the window. I no longer dreaded PRs.
At my current company it's actually a fight as it's not necessarily the architects themselves that want to shit out diagrams all day.
It's a part of the job, for sure, I'm just pointing out it's not always the architect.
but your points about hands-on, respect simple, and having a wide base of experience is 100% spot-on imo. I do agree with your point about the diagrams, just wanted to point out that's sometimes due to pressures rather than wants.
Outside of actual system design and writing code, I have spent a huge amount of my time over the last few years doing things like setting up CI/CD, E2E testing, writing templates, etc. Because it’s really hard to get this right, to have a vision for how things should be.
My point is that it sounds like you’re doing an awesome job. I’m sure you qualify as a software architect.
I generally try to keep design out of architecture. It’s what you need to satisfy product or service needs. The technology gets into how you want to implement. I try to leave that to the tech lead for the team.
It's a common fallacy to assign 50/50 probability to unknown distributions. Furthermore, many choices are not binary, or may not be as simple as right/wrong in hindsight.
But I agree with the idea of amortizing your perspective over many choices.
I usually default to some large company's style guide so it's difficult to argue against. For example, for .NET we use Microsoft's style guide for React we use Facebook's, etc.
That was true for jQuery as well; and yet, it is clearly no longer the right thing (tm) to choose for a project. Not because it's no longer cool; but because the underlying technology — the web browser — has gotten so much better it is no longer relevant.
Same fate may await React. At least, with jQuery it's pretty straightforward to gradually wean oneself off of it. I don't know how firmly we are going to be stuck with React when its time comes.
It was that someone put in the time to write a JS preprocessor that went along with their framework. That was React’s novel addition
Without a preprocessor, users of your framework either use plain JS to generate HTML (i.e. jQuery) or you use templates.
Both of them always looked really ugly on frontend
With React’s JSX preprocessor, you can write HTML by… writing HTML. And you can mix custom components without doing some weird and ugly syntax
Does React use a shadow DOM? Yes, but so could have jQuery or Handlebars.js too, and it wouldn’t have mattered. People don’t care how it works as long as it works. As far as libraries go, what gets adopted is what looks and feels the nicest — as it should be
No. You must be confusing it with virtual DOM.
As for React, what we are currently observing is that 1) it's a fairly large library in itself (~50kB of minified gzipped javascript); 2) it has a relatively slow DOM update mechanism; 3) it rerenders too eagerly, making it too easy to further degrade performance; 4) it's a lock-in into a non-standard jsx syntax, forever tying you to preprocessors.
What conclusions an architect would make looking at all this, I do not know.
The concern is abandoned projects that no one knows how to use, not that something is a bit old. React is beyond that point. It's not just the tech junkies hopping from the latest fad using it. There's been enough widespread adoption that there will still be React projects in 20 years.
[0] https://www.econtalk.org/annie-duke-on-the-power-of-quitting...
- I have a great time listening to "Software Unscripted" podcast.
- On youtube the channel "code opinion". Although I don't like or use DDD and C#, I think that it has high quality and rewarding content.
- I am currently finishing "A Philosophy of Software Design" by John Ousterhout.
- Regarding articles, I read stuff related to my day's obsession or problem. I can't point you out to one. I mean, what do you need to evaluate/learn?
If you need something more specific topic-wise I am glad to share with you
Really enjoying the Code Opinion videos so far.
>"Trying to persuade people to avoid technical debt at all costs" - put in a linter
This for example does not make sense to me. This would mean all tech debt is just static analysis. There is no linter for figuring out the correct abstraction and using them correctly. Eg. no linter yet will tell you "hey you should have used a state machine here!"
You need a width range of experience before you can develop "better" taste in what's good and bad in different scenarios. Some of it can be gained from years of experience, and some from exposure to books/code.
>in my experience
Friend, you're making it clear that you don't have much.
* Yes, it is better that they articulate what they don't like than just completely ignore it and leave me in the dark.