[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gnu-arch-users] Archives vs. categories vs. versions

From: John A Meinel
Subject: Re: [Gnu-arch-users] Archives vs. categories vs. versions
Date: Mon, 15 Nov 2004 22:54:40 -0600
User-agent: Mozilla Thunderbird 0.9 (Windows/20041103)

Dimitrie O. Paun wrote:
On Tue, Nov 16, 2004 at 09:56:11AM +1100, address@hidden wrote:


Why the time marking? I understand it's needed to work around
some performance issues, but in all honesty it looks like a
hack. Moreover, it's hard to automate. Why is it needed in to
begin with?

As far as I know, it is solely for performance. The problem is that after you get 200+ revisions into a single version, you start noticing that some of the algorithms are O(N) (example, checking to see if you have an old revision in your library when the lib is empty causes a round trip for each revision, Cycling your archive breaks this check).

It isn't truly necessary. I'm not sure exactly what happens if you bump the version. I think the search crosses version/category boundaries, but not archives.

I think it would also be reasonable to tell the revlib searching code "if you haven't found it in N steps, just assume it's not there." N could be 10, or it could be 50, either way would probably help.

For projects within a big organization you might have the mainline trees in an archive named after the app, because
that's not going to change, but otherwise I suspect it just
hinders things.

What would it hinder? And what about large sites, say SourceForge?
What if they switch to arch? It would seem quite natural to me
to create an archive for every project:
Having the year in there would create lots of trouble and maintenance
costs going forward. Can it be avoided? And at what cost?

I would generally say you get a full archive depending on how much turnover there is of your source, otherwise just put everything into one. I don't really know what that number is. But for a business' repository, it would make sense to try and keep the number low. Though if you really got into arch, everyone would probably have their own archive, and you would just have the main archive as a centralization point.

I certainly find that distributed archives work very well. (How else would I be helping the development of arch without central commit rights.)


Attachment: signature.asc
Description: OpenPGP digital signature

reply via email to

[Prev in Thread] Current Thread [Next in Thread]