D. Van Buren, R. Ebert, D. Egret
We'll give a brief description of our server. A more detailed version
will appear in the AJ. Our design goal was to give users a simple interface
to the ISSA so both casual browsing and more heavy-duty "survey" mode data
gathering would be easy. An important principle was that we would provide
the ISSA data with no further processing - this was to be a way to get at the
released data, not a way to reformat, rebin, reproject etc. Those tasks are
all better left to the user's software. Obviously, reformating, rebinning and
reprojecting all change the data so it is no longer the ISSA data but something
else. The second important principle was to design the server so that other
programs could make a TCP/IP connection, make a request, and then get back
enough information to pick up the fits images automatically. In this way we
can support large surveys, interfaces from other programs etc. It turned out
that it was simple just to implement the server as an html server and then
remote programs only need to understand enough html to get the information
they need. This way NCSA Mosaic is a convenient browsing interface.
We only ship one size image - 2 degrees by 2 degrees. This is the
largest image we can make that is guaranteed not to require mosaicking
adjacent plates. Those who need larger images can ftp the entire plates
from the on-line archive.
When users dial in to http://brando.ipac.caltech.edu:8888/ISSA-PS
they connect with a MOO server which supports a large number of interfaces
to network data services, and allows synergistic interactions between them.
The MOO server interprets the web request, formats it and dispatches it to
the core postage stamp server (which does not speak html). If the request
is by name, it first makes a position-by-name request to SIMBAD. The core
server has several stages. It looks up the ISSA plate containing the
requested field based on the coordinate. Then it snips out the four subimages
corresponding to each of the four IRAS bands from the ISSA dataset on CDROM.
These are placed in a local public ftp area. At the same time it converts
each of these to a gif image for use by web browsers and places them in the
ftp area as well. A list of created files is then sent back to the MOO server
where an html document is created inlining the gifs and pointing to the fits
files. The html is then sent to the user. An important part of the service
is a "Read This!" section which lists caveats and other need-to-know items
about the ISSA images.
The postage stamp server query page has instructions how to configure
NCSA Mosaic to retrieve fits files as binaries. For small requests users
can click on the inlined images, but for large surveys they would likely find
a survey engine easier to use. We have been playing around with this inside
the MOO server, where one can currently edit a list of target names and get
the ISSA postage stamps dumped on local disk.
The service has been running since December 1993 and averages about
ten requests a day. In the old days before ISSA, when we made "BigMaps" from
the IRAS data almost by hand, it cost about 2 FTEs to service the same request
rate. With the ramp down of IRAS activities at IPAC, the postage stamp server
is the only way we can inexpensively make this data publicly available to all
comers.