[EMAIL PROTECTED] on 02/28/2000 10:01:35 PM
>[ On Monday, February 28, 2000 at 16:57:30 (-0500), Noel L Yap wrote: ]
>> Subject: Re: removing the need for "cvs add file" to contact the server....
>>
>> Again, you're absolutely right. Users shouldn't rely on the optimisation you
>> propose. They should instead be using the ignore facility.
>
>I think you've totally missed the significance of a very major point.
>All of your argument crumbles to a great void in the face of it.
>
>That point is that the CVS "ignore" facility will be over-ridden by any
>explicit specification of any name on the command line that would be
>otherwise ignored. This goes for both directories and filenames.
You're absolutely correct again. If someone has a .cvsignore file listing all
their empty directories and does "cvs add; cvs ci", those empty directories are
ignored. If someone does a "cvs add empty-hier; cvs ci", he has explicitly said
to add the empty directories. The tool should not prevent him from doing so,
specially if it only does so as an "optimisation". Your argument that the
NetBSD people will appreciate this optimisation has no foundation since they can
easily achieve what you describe by:
1. using .cvsignore
2. not specifying the empty hierarchy as an argument to "cvs add"
>> Furthermore, your optimisation *will* cause CVS to do extra work when it
comes
>> to empty directory hierarchies.
>
>Hmm... really? Are you really sure you know what kinds of
>optimisations I'm talking about?
Yes, one of my responses was to give pseudo-code to the algorithm used. You
responded affirmatively. The algorithm has a lot of file system access.
In the same post, I also gave pseudo-code for the algorithm without the
"optimisation". The complexity is a lot less without the optimisation.
>Since those optimisations are already a part of the existing
>implementation perhaps you can show us with concrete and tested examples
>of how you think they could cause CVS to do extra work on empty
>directory hierarchies.
I already did. If you want, I'll dig through my archives to post them again
(I'll even post your response to them). The alternative is that you dig through
the archives since you probably won't believe what I post anyway.
>> This is not a side issue. CVS commands are CVS commands. You cannot be
biased
>> towards/against one "set" of commands just 'cos you don't use them.
>
>On the contrary.
>
>But regardless the solution for them will be fairly obvious once the
>rest of it is worked out....
So long as this issue is dealt with, I'm fine with it (but not the
"optimisation").
Noel
