Code review comment for lp:~rls/bzr/transport-link-and-symlink-support

> However it seems like paramiko normally translates errors into
> exceptions itself, so you may not need to worry about this?

Not sure about that part. I think it raises errors on blatant errors (disconnection, non-support-- stuff that the library can't handle without human intervention, I suppose). For the most part, it returns error codes. Crude, but hey, I just work with it. :P

> Actually the thread seemed to conclude that it was better for Transport.stat()
> to call os.lstat, and then make the higher level code specifically deal with
> the symlink.

Okay, I've reverted that, then.

> Like that, except the osutils.hardlink_good() check seems redundant. Just
> declare the method always and let the AttributeError call handle the case that
> they're not supported.

Eh. Your suggestion conflicts with gz's. I do like the idea of letting the rest of bzrlib's guts tell me whether it's okay to declare a method, though, rather than handle an AttributeError myself. But you guys are more familiar with the gestalt of Bzr than I am, so perhaps I'm missing something?