[Clipart] storing hashes

[Jon Phillips]

On Wed, 2005-08-17 at 13:42 -0400, Greg Steffensen wrote:
> I'm working on the OCAL interface for Inkscape, and am wondering if a
> specific addition to the DMS api would be possible to accomodate it.
> When I proposed implementing this thing, multiple people suggested
> that I make it easy for the interface to access clipart repositories
> other than OCAL proper, including some kind of shared system
> repository at /usr/share/clipart (on Linux).  In brief, I implemented
> this by having the interface search a given list of repositories in
> sequence, and if a particular image is available from more than one
> repository, its retrieved only from the first repository on the list
> (this is supposed to be vaguely analagous to the multi-tiered caching
> systems processors use to access memory).  
> 
> This requires some way to identify images uniquely though, so that
> duplicates can be detected, and for lack of a better alternative, I'm
> using the md5 hash of the svg xml contents.  I have this working
> correctly for a local repository as described above, but would like to
> know if the eventual SOAP (or whatever) api intended for programmatic
> access could return, in addition to a list of image ids of some sort,
> their hashes.  

I think this sounds wise. This would be useful as well across the wires
for identifying files. Bryce, what do you think?

> I've made this optional in the interface... if a repository doesn't
> return hashes, then whatever images it finds are just retrieved.  Of
> course, I could calculate the md5 hashes dynamically after retrieval
> to prevent duplicates from being displayed, but that would still waste
> OCAL's bandwidth and the user's time.  So I'm wondering if this scheme
> seems reasonable, and whether OCAL could support it.  Any ideas?

Greg, this seems like a sensible approach to me. Jonadab, could any of
this be rolled into what you are working on for indexing the clip art
collection?

Jon

-- 
Jon Phillips

San Francisco, CA
USA PH 510.499.0894
jon at rejon.org
http://www.rejon.org

MSN, AIM, Yahoo Chat: kidproto
Jabber Chat: rejon at gristle.org
IRC: rejon at irc.freenode.net

Inkscape (http://inkscape.org)
Open Clip Art Library (www.openclipart.org)