Actually one of the most reflected opinions regarding this topic I've seen so far. [Imho] Both approaches can become doomed, and both can lead to great success-stories. In my experience the biggest benefit of small independent repos is their atonomy, their biggest downside is their tendency to create dependency-hell (or it's weaker cousin, orphaned projects, which are still in use). On the other hand the biggest strength of monorepos is their ability to show the whole picture, unfortunately they tend to rely on heavy tooling as their biggest downside. [/imho] Unfortunately I see a trend in the last 5 years, especially in the JS/TS community, to build tooling which only serves a their language and don't support propper support for other tool-chains. This is really bad for monorepos, since one of the key-ideas of such a setup is to not constrain you too much on the language. (Buck/Bazel/Gradle/Nx [mostly language independent] vs. TurboRepo/Lerna [heavily relying on one language]). Another pain-point can be to maintain languages which (by default) rely on a specific structure of repositories or directories, like golang...
My main issue with Monorepos is the required config. If you can't understand the config from a quick-glance, get rid of it! Because when the author leaves the team, you're screwed. Web development is 50% config hacking these days
i just studied nx and turborepo, migrating it require me to study specific configuration for each, and doesnt work at first try, definetly learning curve
I used Next with Nx and it wasn't that bad, to be honest. Worked well with Vercel, as well. Also, people shouldn't own projects, applications, or features. These should be documented so they can be shared and picked up by other Devs in the case that someone leaves, it can be easily picked up.
Its good advice to avoid multiple teams working on a single codebase whether its a monorepo or not. It doesn’t mean that members of other teams can’t contribute, but there should be a single owner of that repo to help prevent disruptive changes to the code. I would consider this a facet of Conway’s Law (the creator touched on this but didn’t name it): development is most efficient when code is structured similarly to the organization’s structure and vice versa. Having a single team own a single repo is more efficient than a single team owning multiple repos or multiple teams owning a single repo. Its not a hard and fast rule, but its been mostly the case in my experience. Anyway, i’ve been using monorepos for years, and I’ve found that its best when only one team works on it. When we get too big we just break the monorepo up as needed (actually in the process of this now after about 14 months on a project). So its just a guideline, but i do agree with the creator’s advice on this one.
Isn't a monorepo for team a separate issue that team can decide without other teams that make direct updates to included submodules even needing to know about it? In other words, a monorepo is an extra aggregating repo that simply allows you to deal with multiple independent repos together. So it doesn't destroy anything by making it available. Am I missing something?
How can I configure the mono repository to use same package in each and every project ? Let’s say we have our UI and API projects in mono repo, both written in JS. Now how can we make both to use same package version ? Because each of them have it’s own package.json to manage its own dependencies, right ?
thanks man you explain very well, but I have a small question what if I have to deploy multiple services with mono repo using CICD, should I wait till a service is done and then execute the next one or is there a way to deploy them at same time?
That's a really good question. It depends on your requirements: do your services depend on each other for ex. is there a specific setup that has to be finished or maybe some library is being used from a parent service? If there is no dependency then you can deploy them in parallel, otherwise you have to configure your pipelines to do it sequentially.
I've been wondering about that too. I think you should be able to set up your CI/CD jobs in such a way that allows for parallel execution. If there are any dependencies, you can set up a job that runs only when the dependency is done and ready. Idk but this is what makes sense to me
Polyrepos are also helpful when you want your code (business) logic to be packed differently for deloyment over to varied hostings over different geographies, as separating the build in other repos works well when working with multiple teams, in a large organization. I have seen people trying to forcefully fit cases in monorepos which is not a good practice. It creates overcomplications and takes things away from the 'Single Responsibility' based thought processes. Just because Google or some other big org does it, shouldn't be the reason to force monorepos everywhere.
part of the gig is finding ergonomic ways to do things...generally what I see is that yes people sometimes "Force" things, but that was not the only option.
Great Job @sofwaredeveloperDaiaries. please i will like to connect with you regarding this topic of mono repo, if you do not mind please.. i have a project there, which i would like to understand somethings
Actually one of the most reflected opinions regarding this topic I've seen so far.
[Imho] Both approaches can become doomed, and both can lead to great success-stories. In my experience the biggest benefit of small independent repos is their atonomy, their biggest downside is their tendency to create dependency-hell (or it's weaker cousin, orphaned projects, which are still in use). On the other hand the biggest strength of monorepos is their ability to show the whole picture, unfortunately they tend to rely on heavy tooling as their biggest downside. [/imho]
Unfortunately I see a trend in the last 5 years, especially in the JS/TS community, to build tooling which only serves a their language and don't support propper support for other tool-chains. This is really bad for monorepos, since one of the key-ideas of such a setup is to not constrain you too much on the language. (Buck/Bazel/Gradle/Nx [mostly language independent] vs. TurboRepo/Lerna [heavily relying on one language]). Another pain-point can be to maintain languages which (by default) rely on a specific structure of repositories or directories, like golang...
Thanks. Please keep posting more such content.
Will do, my friend! Appreciate the Super Thanks! 😊
My main issue with Monorepos is the required config. If you can't understand the config from a quick-glance, get rid of it! Because when the author leaves the team, you're screwed. Web development is 50% config hacking these days
Good point
i just studied nx and turborepo, migrating it require me to study specific configuration for each, and doesnt work at first try, definetly learning curve
I used Next with Nx and it wasn't that bad, to be honest. Worked well with Vercel, as well.
Also, people shouldn't own projects, applications, or features. These should be documented so they can be shared and picked up by other Devs in the case that someone leaves, it can be easily picked up.
@@incarnateTheGreat Agree 100% with that last statement.
Once a while the UA-cam algorithm recommends a great channel. Now is such a time.
Nice to meet you - keep posting great content
Nice to meet you too! Thanks for the kind words 😊
Why only 1 team on a monorepo. I would assume we just resolve the merge conflicts with the clashing team? What do you think?
no more than one team use monorepo, what??? Super confused. Are you suggesting multi-repo instead of MONOrepo? How can a company only has one team???
I’m at a start up, we’re just 5 devs i.e. a single team
@@gnrsn9714Sure, but limiting a monorepo to one team as a rule contradicts one of the main points of monorepos.
Its good advice to avoid multiple teams working on a single codebase whether its a monorepo or not. It doesn’t mean that members of other teams can’t contribute, but there should be a single owner of that repo to help prevent disruptive changes to the code.
I would consider this a facet of Conway’s Law (the creator touched on this but didn’t name it): development is most efficient when code is structured similarly to the organization’s structure and vice versa. Having a single team own a single repo is more efficient than a single team owning multiple repos or multiple teams owning a single repo. Its not a hard and fast rule, but its been mostly the case in my experience.
Anyway, i’ve been using monorepos for years, and I’ve found that its best when only one team works on it. When we get too big we just break the monorepo up as needed (actually in the process of this now after about 14 months on a project). So its just a guideline, but i do agree with the creator’s advice on this one.
Isn't a monorepo for team a separate issue that team can decide without other teams that make direct updates to included submodules even needing to know about it? In other words, a monorepo is an extra aggregating repo that simply allows you to deal with multiple independent repos together. So it doesn't destroy anything by making it available. Am I missing something?
Not really
How can I configure the mono repository to use same package in each and every project ? Let’s say we have our UI and API projects in mono repo, both written in JS. Now how can we make both to use same package version ? Because each of them have it’s own package.json to manage its own dependencies, right ?
You want to search for workspaces. npm and all the other major package managers have a way to do this.
thanks man you explain very well, but I have a small question what if I have to deploy multiple services with mono repo using CICD, should I wait till a service is done and then execute the next one or is there a way to deploy them at same time?
That's a really good question. It depends on your requirements: do your services depend on each other for ex. is there a specific setup that has to be finished or maybe some library is being used from a parent service? If there is no dependency then you can deploy them in parallel, otherwise you have to configure your pipelines to do it sequentially.
I've been wondering about that too. I think you should be able to set up your CI/CD jobs in such a way that allows for parallel execution. If there are any dependencies, you can set up a job that runs only when the dependency is done and ready. Idk but this is what makes sense to me
Polyrepos are also helpful when you want your code (business) logic to be packed differently for deloyment over to varied hostings over different geographies, as separating the build in other repos works well when working with multiple teams, in a large organization. I have seen people trying to forcefully fit cases in monorepos which is not a good practice. It creates overcomplications and takes things away from the 'Single Responsibility' based thought processes. Just because Google or some other big org does it, shouldn't be the reason to force monorepos everywhere.
part of the gig is finding ergonomic ways to do things...generally what I see is that yes people sometimes "Force" things, but that was not the only option.
we have about 250 repos
Oopw
Great Job @sofwaredeveloperDaiaries. please i will like to connect with you regarding this topic of mono repo, if you do not mind please.. i have a project there, which i would like to understand somethings