carbon-lang icon indicating copy to clipboard operation
carbon-lang copied to clipboard

Expand non-goal section in README

Open rscircus opened this issue 3 years ago • 2 comments
trafficstars

Add non-goals from https://github.com/carbon-language/carbon-lang/blob/trunk/docs/project/goals.md#non-goals

rscircus avatar Jul 25 '22 14:07 rscircus

carry bit from #1671:

@chandlerrc: I don't quite agree with these three.

The first addition I think is a goal, but maybe not the top goal. I think it requires a lot more nuance which the interop strategy gets into and we can't really cover here.

The second one I think is likely to be misunderstood. A goal is to work with existing C++ build systems, and that means we need at least a largely similar compilation and linking model. How much we can diverge there I don't think is clear yet, and so I wouldn't put this as a non-goal. I worry it would be confusing at best. =/

The last one I think is technically true, but likely to be misleading. I feel like the longer goals document talks about this with more of the nuance it needs.

Still, interested in other thoughts about these additions. I think these would at least need to be a proposal as I think it might be a more substantial change that should have a quick rationale and the leads sign off on it. The fix above though should just move forward quickly as a simple wording improvement.

@rscircus

The first addition I think is a goal, but maybe not the top goal.

image

This point is actually verbatim taken from the first sentence, which is the same as in the goals.md document here.

The second one I think is likely to be misunderstood.

Agreed.

The last one I think is technically true, but likely to be misleading.

Alternative suggestion:

->                     Stringently support legacy libraries for which source code is unavailable
-   Prioritize compilation and linking model over needs of C++
-   Adhere to or reproduce all C++ semantics

rscircus avatar Jul 25 '22 19:07 rscircus

While I agree with keeping those documents in sync, I would prefer using the exact wording from the goals document.

Initially, I thought the same as you did, @ooxi. But then there is Feynman... and I thought I do not follow the academically true way but weaken it up a bit to make it understandable as it's a README and not a RFC or proposal. Hence, also the wikilink addition around ABI in #1671...

rscircus avatar Jul 25 '22 19:07 rscircus

Closing this/my PR as it has too much potential for getting discussed to death costing valuable lifetime without adding value for a long time.

rscircus avatar Sep 10 '22 07:09 rscircus