Testing an MIT-wide cache to speed web access

If you visit the same web sites repeatedly, or if you visit highly-popular web sites, you may find that things speed up a lot when you try this experimental cache. The only catch is that if the service goes down, your web browser will give an error message and not retrieve any documents. So make sure you know how to turn it off before you turn it on.

Here are intructions on starting/stopping testing with various WWW browsers:

most Unix browsers, including Mosaic and lynx
Netscape
IBM's WebExplorer

Technical details

This cache is hooked into the NLANR cache project, using software called Squid, which is faster than other web cache software. To find out why, read about the Harvest Object Cache, which Squid is based on.

Platform:	100MHz SGI Indy running IRIX 5.3
Memory:		16M cache on 64M machine
Disk:		750M cache on 1G Fujitsu M2694ES

Limited use outside of MIT

Some servers restrict access based on hostname or IP address, e.g. the servers for Publications Licensed to MIT. When you use a proxy server, the IP address the server sees is that of the proxy. For this reason, access to those servers via the proxy on squid.mit.edu is restricted to MIT.

Some servers at MIT may also have pages that are hostname or IP-restricted. For this reason, access to MIT servers is limited to those on my list of servers known to have no such restricted pages. Other access restrictions (e.g. password-based) are ok. If you know of an MIT server that should be added to this list, please let me know.