Re: [stellation-res] Adding artifact priority

["Mark C. Chu-Carroll" <mcc@xxxxxxxxxxxxxx>]

On Saturday 10 August 2002 03:24 am, Florin Iucha wrote:
> On Fri, Aug 09, 2002 at 09:27:15PM +0000, Mark C. Chu-Carroll wrote:
> > > All that really matters for an initial test is that Compound
> > > representing a directory (as all Compounds currently do) have a
> > > distinguished priority.  I would suggest default priority of 0, with
> > > Compound having priority 1.
> >
> > I'd like to suggest doing something slightly different. I'd like the
> > option of putting things at both lower and higher priorities than
> > compounds. So I'd like to select a range for priorities - say 1-200. Then
> > put compounds right in the middle.  Things that need to run before
> > compounds
> > get priorities less than 100; things that run after the compounds
> > get priorities greater than 100. At the same priority, we don't
> > define the order. I'm not sure how large a priority range we want. I
> > think 200 should be plenty - I can't imagine 200 artifact agents! - but
> > people who make assumptions like that often get quoted years later
> > for their foolishness :-).
> >
> > Actually, how about just make the priorities an integer, let them be
> > signed, and run them in increasing order? Then we put Compound at 0,
> > which is a nice central distinguished position. That way, we've got as
> > large a range as we could possibly want.
>
> I think you need an extra bit to specify if the compound is ordered or
> not. The order of files in a directory is irrelevant. The order of lines
> in a source file is relevant.

I think we're tripping over another bug in the docs. (Don't worry, I
am keeping track of the doc changes we need to make, and they
will get done when one of us has a free moment!)

A compound is repository-speak for a directory. When we first
started, before we even came up with the agent notion, we had two
kinds of artifacts: Atomic and Compound.

Originally, when we were planning for fine-grained support, we thought that 
compounds would be used both for representing directories made from 
collections of files, and other kinds of aggregate types like the collection 
of fine-grained elements that made up a file.

Over time, we realized that directories and fine-grained aggregates
were very different things, and it didn't make sense to use compounds
for both.

The (not yet implemented) artifact type for non-directory aggregates
is designed to support ordered and unordered collection types. (For
details, take a look at the paper in our publications section. It's the
FSE2002 preprint.

	-Mark

-- 
Mark Craig Chu-Carroll,  IBM T.J. Watson Research Center  
*** The Stellation project: Advanced SCM for Collaboration
***		http://www.eclipse.org/stellation
*** Work Email: mcc@xxxxxxxxxxxxxx  ------- Personal Email: markcc@xxxxxxxxxxx