On Tue, 27 Jan 2004, Marco Antoniotti wrote:
> I think CMUCL does this. You can check the past mails. I do not know
> if this has been ported to SBCL.
> If I remember correctly it does
IIRC yes. But the downside is that you get something that to me at least
seems inconsistent and hard to customize:
foo.ext~42~ => type ext, version 42
foo.ext~1.4.2~ => type ext~1.4.2~
OTOH, a "low cost" "improvement" to pathnames might involve making the
backup suffix customizable, and add symbolic version :backup (actually
allowed by the standard) that can be used to retrieve the backup version.
So you could
(setf sb-ext:*backup-suffix* "~")
to use emacs style non-versioning backups.
Then again, namestring parsing could not really be smart about this, as
then parsing "foo.ext~42~" would now result in type ext~42, version
I think the moral here is that unix pathnames and type/version suffixes
are way too undefined to sanely map to CL pathnames in any other way then
"everything after the first dot is type", or "no version, no type, just
name". Otherwise a user is going to in the end lose due to the mismatch.
Just thinking out loud,