5 ms·

And IMO that's a really good way of solving this issue! Rather than try to push a paradigm shift under a major version change in a library, and rather than jus

And IMO that's a really good way of solving this issue! Rather than try to push a paradigm shift under a major version change in a library, and rather than just hand the library over to someone who may or may not be vetted enough, they are gracefully shutting it down while still sticking around to fix any security issues that may crop up.

chadlavi 7y ago
This is what semantic versioning is for, though. Breaking changes to the API that still serve the same function in an app would not unreasonably be a new major version of the same package. No one is obligated to upgrade to the latest major version. A complete rewrite would be a ship of theseus situation. It's all new code, and the API may be different, but it still exists to serve the same purpose in an app. Why pretend it's a new module when it's really just a retooled version of the same one? EDIT: Ok, I misunderstood what was going on with the request maintainers; thought they were planning to start up a new package as a rewrite of request. That practice specifically is what I was arguing against here -- it's not good practice to start a new node module with a new name just because you have breaking changes to your API. I do not in any way mean to suggest that someone else should jump in and turn request into something new.
- Klathmon 7y ago
  The author of that issue linked above says it better than I ever could: > A version of request written to truly embrace these new language patterns is, effectively, a new module. I’ve explored this space a bit already and have a project I’m quite happy with but it is incompatible with request in every conceivable way. What’s the value in a version of request that is incompatible with the old patterns yet not fully embracing the new ones? What’s the point in being partially compatible when there’s a whole world of new modules, written by new developers, that are re-thinking these problems with these patterns in mind? > The best thing for these new modules is for request to slowly fade away, eventually becoming just another memory of that legacy stack. Taking the position request has now and leveraging it for a bigger share of the next generation of developers would be a disservice to those developers as it would drive them away from better modules that don’t have the burden of request’s history. Now view that in contrast to react-router, which has gone through 3 or 4 really major rewrites, and had their API change those few times in completely incompatible ways, even shifting paradigms multiple times over the course of the library until now. They are still committed to maintaining most of the past major branches, and the docs are still around for those versions, but each one is so different that it adds a lot of confusion. And I don't meant to say that react-router is bad, the reasons for the changes are damn good ones, and I actually like the library a lot, but the confusion around it is real because the 2.x branch vs the 3.x branch vs the current branch are all so different that they should really be different libraries in my opinion. If I had to choose between those 2 things, i'd choose what request is doing both as a library author and a library user.
  - chadlavi 7y ago
    I misunderstood -- I thought the authors of request were going to keep working on a rewritten version, but were going to start a brand new node module for it. Strictly speaking, react-router is doing it correctly. It's the same library, just different versions. The normative way in the npm community to handle breaking API changes is to increase the major version, not to completely rename the package.
- cipherboy 7y ago
  Yeah but think of it from an (Linux OS) maintainer perspective: if you don't have a supported way to install parallel versions of the same package, you're stuck creating it under a new name (e.g., request999) because packages won't update. This is the solution in, e.g., Fedora where "junit" means junit4 and "junit5" is a separate package. Over time, junit might become deprecated and be removed from Fedora when all deps have moved on and junit5 takes over the junit package name, perhaps a junit6 will emerge. Or junit4 gets recreated for the last two legacy apps using junit to live on. Who knows! But from a maintainer perspective, better to have a separate name than a new breaking version... :-)
  - chadlavi 7y ago
    Who said anything about linux packages? this is NPM. I'm a library maintainer, and it doesn't make any sense to me to create totally different packages with different names just to handle a breaking change. This is what semantic versioning exists for.

chadlavi 7y ago

This is what semantic versioning is for, though. Breaking changes to the API that still serve the same function in an app would not unreasonably be a new major version of the same package. No one is obligated to upgrade to the latest major version. A complete rewrite would be a ship of theseus situation. It's all new code, and the API may be different, but it still exists to serve the same purpose in an app. Why pretend it's a new module when it's really just a retooled version of the same one? EDIT: Ok, I misunderstood what was going on with the request maintainers; thought they were planning to start up a new package as a rewrite of request. That practice specifically is what I was arguing against here -- it's not good practice to start a new node module with a new name just because you have breaking changes to your API. I do not in any way mean to suggest that someone else should jump in and turn request into something new.

Klathmon 7y ago
The author of that issue linked above says it better than I ever could: > A version of request written to truly embrace these new language patterns is, effectively, a new module. I’ve explored this space a bit already and have a project I’m quite happy with but it is incompatible with request in every conceivable way. What’s the value in a version of request that is incompatible with the old patterns yet not fully embracing the new ones? What’s the point in being partially compatible when there’s a whole world of new modules, written by new developers, that are re-thinking these problems with these patterns in mind? > The best thing for these new modules is for request to slowly fade away, eventually becoming just another memory of that legacy stack. Taking the position request has now and leveraging it for a bigger share of the next generation of developers would be a disservice to those developers as it would drive them away from better modules that don’t have the burden of request’s history. Now view that in contrast to react-router, which has gone through 3 or 4 really major rewrites, and had their API change those few times in completely incompatible ways, even shifting paradigms multiple times over the course of the library until now. They are still committed to maintaining most of the past major branches, and the docs are still around for those versions, but each one is so different that it adds a lot of confusion. And I don't meant to say that react-router is bad, the reasons for the changes are damn good ones, and I actually like the library a lot, but the confusion around it is real because the 2.x branch vs the 3.x branch vs the current branch are all so different that they should really be different libraries in my opinion. If I had to choose between those 2 things, i'd choose what request is doing both as a library author and a library user.
- chadlavi 7y ago
  I misunderstood -- I thought the authors of request were going to keep working on a rewritten version, but were going to start a brand new node module for it. Strictly speaking, react-router is doing it correctly. It's the same library, just different versions. The normative way in the npm community to handle breaking API changes is to increase the major version, not to completely rename the package.
cipherboy 7y ago
Yeah but think of it from an (Linux OS) maintainer perspective: if you don't have a supported way to install parallel versions of the same package, you're stuck creating it under a new name (e.g., request999) because packages won't update. This is the solution in, e.g., Fedora where "junit" means junit4 and "junit5" is a separate package. Over time, junit might become deprecated and be removed from Fedora when all deps have moved on and junit5 takes over the junit package name, perhaps a junit6 will emerge. Or junit4 gets recreated for the last two legacy apps using junit to live on. Who knows! But from a maintainer perspective, better to have a separate name than a new breaking version... :-)
- chadlavi 7y ago
  Who said anything about linux packages? this is NPM. I'm a library maintainer, and it doesn't make any sense to me to create totally different packages with different names just to handle a breaking change. This is what semantic versioning exists for.