|
Progress report: Image Server/MRML connection software
|
Wolfgang Müller
|
Oct 19, 2001 02:47 PDT
|
On Thursday 18 October 2001 14:16, you wrote:
| | Dear Benchathletes,
* Concretising the benchmark architecture: The idea is define the
harness to include any system within our framework. Wolfgang and Neil
have been working on that. What is the status right now (Wolfgang, a
short description and tutorial on the harness?)?
|
OK. Here I will talk rather about the Image Retrieval server side of things.
I have been rather busy with my thesis, but I have thought a bit about the
issues Neil had raised. The problem he saw is, that as it currently is, my
Perl example MRML server is too difficult to install for anyone who is not a
Perl fan. I saw this point, especially as I found an unreported bug when
installing its Windows version with him looking over my shoulder.
--> I guess the easiest and most platform independent way is to package the
derver and send each new package version to CPAN (http://www.CPAN.org). CPAN
provides the standard installation mechanism of Perl (*very* simple to use),
and makes sure that prerequisites are fulfilled on installation. It's more
administration overhead for us (I am currently learning that), but I guess it
will be worth it. When the first version is on CPAN, I will drop a note.
For making development more transparent, I will put the current version of
the example server on CVS at SourceForge.
The second issue Neil raised was that you have to know object oriented
features of Perl to make use of it. Well, most people don't, so I agree with
him, that I need to create some callback function-oriented interface to the
perl-example server.
--
The last and most important suggestions, was to create something simple for
very specialised people who don't want to hack (nor program), and who have
nothing but a command-line tool, and who do not want (read: they do not have
the resources) to learn Perl/Shell just for the Benchathlon. He gave a graph
server as example, whose scheme we cannot use, as the human user has the role
of a *client* there, and not the role of the server. To get around this
tricky problem, I suggest the following solution:
Let's assume that the user has at least a shell (or command line). We could,
instead of sending a query, just send him (or her, I will use "him" as short
for "him or her" in the rest of this mail, please excuse) a mail that has a
shell script attached. The user executes the shell script, and redirects the
output to a file. We send him a new mail, with our feedback query, and so
forth.
Note that this is only feasible if we can process batches, and build a file
with several retrieval results. While I think this might be worthwile trying
(though *hard* to debug), I need some information from you:
--
What do you use as a tool, do you use a command line interface, what does
your command line look like, what does the output look like?
--
I am hoping for explanations and subsequent vivid discussions.
| | It seems relatively critical that something from the Benchathlon is
demonstrated at SPIE.
|
So in this sense, I suggest the following sequence of events (from my side):
1) perl-example-server -> CVS
2) Changements on perl-example-server
-> use of standard Perl XML tree representation instead of homebrewed
(for easier distribution)
-> Callback mechanism
-> CPAN package
--here ends the stuff I can guarantee till end of the year
3) Mailbased solution (if feedback, and if interest)
For testing 2) we would need some people with very little Perl experience, if
possible also some Windows users.
Comments, corrections and extension welcome
Cheers,
Wolfgang
--
Dr. Wolfgang Müller, assistant == teaching assistant
Personal page: http://cui.unige.ch/~vision/members/WolfgangMueller.html
Maintainer, GNU Image Finding Tool (http://www.gnu.org/software/gift)
|
|
 |
|
