Fwd: [Openmcl-devel] Possible problem with host-namestring

[Gary Byers]

On Sun, 4 Jan 2004, Gary King wrote:

> I think that the problem indicated below is legitimate. Is there a
> reason for OpenMCL's differing behavior (that I will work around in my
> own code) or is OpenMCL in the "wrong".

One reason is "whim": the namestring returned by HOST-NAMESTRING is
inherently implementation-dependent, and there are lots of differences
between the pathname implementations of MCL and OpenMCL.

One reason I can see for OpenMCL's HOST-NAMESTRING tacking a
colon onto the name of a defined logical host is to slightly simplify
the implementation of NAMESTRING.  Having HOST-NAMESTRING behave the
way that it does is that the behavior of:

? (pathname-host (host-namestring "HOME:"))

is arguably more consistent and intuitive, though using terms like
"consistent and intuitive" when discussing pathnames is asking for
trouble.

Those may or may not seem like compelling reasons for differing from
MCL's behavior, but it's certainly not wrong for implementations to do
implementation-dependent things in implementation-dependent ways.

It's sometimes hard to avoid writing code that depends on implementation
arcana (this is perhaps especially true when pathnames are involved), but
if it's possible to do so in your case it's probably desirable.

> Thanks in advance.