[1567] in Moira

home help back first fref pref prev next nref lref last post

Re: access_list() fixes

daemon@ATHENA.MIT.EDU (Dan Winship)
Fri Jun 2 19:39:48 2000

Message-Id: <200006022339.TAA03133@technomage.mit.edu>
To: Garry Zacheiss <zacheiss@MIT.EDU>
Cc: Jonathon Weiss <jweiss@MIT.EDU>, moiradev@MIT.EDU
In-Reply-To: Your message of "Fri, 02 Jun 2000 05:29:37 EDT."
             <200006020929.FAA14289@shock-treatment.mit.edu> 
Date: Fri, 02 Jun 2000 19:39:44 -0400
From: Dan Winship <danw@MIT.EDU>

>    The client will print "Unknown error code 71" or whatever like any
> com_err application would.  This shouldn't stop us from adding new error
> codes, since the logical conclusion is that we should never add another
> error code again.  People with local moira clients earn no pity if
> they're not clueful enough to cope.

Don't forget about people with Windows and eventually Mac clients that
aren't maintained from the moira locker.

Or the web-based moira client.

As for "we should never add another error code again", you can
actually add new error codes without guilt if you can guarantee that
old clients will never see them. Eg, MR_VERSION_HIGH and
MR_VERSION_LOW, the only two new errno values I ever added, which can
only be returned by a client new enough to know about query
versioning.

So one happy way to fix this would be to bump the version number on
ulis but not actually change the interfaces, and only return the new
code if the client uses the higher version number, and use MR_PERM
otherwise. That might be overkill though.

-- Dan

home help back first fref pref prev next nref lref last post