Timeline to remove process.binding debt from core?

[bmeck]

There is an odd comment about readable-stream requiring a call to process.binding, but in a cursory glance on this repo I don't see usage that gets access to the native source text to evaluate.

https://github.com/nodejs/node/blob/63b056d8d4f0696254cd5fc40a69aee0157fc410/lib/stream.js#L57

Is there some timeline to remove this code / when does the support for ancient forms of readable-stream expire?

[addaleax]

I’m not sure if that’s what you’re asking here, but the code you linked to has an automatic fallback for when process.binding() does not return a valid result. There is no need to remove it, imo.

We should probably undo nodejs/node@c97851dcd8c5e for that file (and the other stream files), though.

[bmeck]

Why/how is that code being used that the binding call needs to remain / streams.js seems to be working even with other internal specific API usage so I'm questioning the actual benefits of the odd fallback

[addaleax]

I don’t really understand what you are saying, sorry. The code is there for the versions exported as readable-stream, and the process.binding call is only used for older versions of Node.js.

[bmeck]

The code is there for the versions exported as readable-stream, and the process.binding call is only used for older versions of Node.js.

This doesn't make sense to me, where is it being used?

[bmeck]

Or to put it another way, why is the binding important, but the other internal only things not? It seems v12 passes tests even with a variety of missing internals due to rewriting source, but it doesn't cover primordials and it still passes. There isn't a disclaimer or anything in that file about it being special and the list of internal APIs being used in it is increasing over time:

v8

• internal/streams/legacy
• internal/util/types w/ fallback via process.binding
• internal/buffer w/ fallback on userland impl

v10 added

• internal/streams/pipeline
• internal/streams/end-of-stream

removed

• internal/util/types

v12 added

• primordials
• internal/util/types again with fallback

[mcollina]

I think we can safely remove that block from core, it's from the old days and for an old line of readable-stream.