IoError is kind of archaic

[bluss]

The documentation for IoError still says # FIXME Is something like this sufficient? It's kind of archaic.

This is an issue to track this untagged FIXME. It should be addressed or the documentation should be updated -- I'm reporting this due to a user question about IoError on IRC.

[aturon]

cc me

[reem]

cc me

[brson]

I may have written the FIXME, and probably I was thinking that it was 'archaic' to bubble up unixy I/O error codes with an ancient legacy and loose mapping to non-unixes.

In hindsight I'm no longer too concerned about that. The names have been mostly well-rustified, and they are abstracted across platform reasonably well by virtue of the I/O stack being implemented cross-platform, and libuv already having done much of the error-unifying work for us.

Another concern is that the list of possible errors is static.

The naming of the variants should be scrutinized.

[reem]

Relevant: rust-lang/rfcs#201

[alexcrichton]

This is quite an old issue and io::Error is now stable, so closing