bajan@cs.mcgill.ca (Alan Emtage) (06/13/91)
As you have probably already seen Clifford has been able to incorporate archie into the Prospero family, and very nicely I may add. I should point out that all the work was done by said Cliff and we just provided the interface routines (which we use internally for the current interactive front end). Having poked around a bit, I have to say I'm quite impressed at how nicely things seemed to have dovetailed although that might be coding magic on Cliff's part. Nice job. Having done this, the first step in splitting archie into a client/server model has been taken. The 'prog' standalone client talks to server on the archie host (via the Prospero protocol) and retrieves the information. As Clifford has mentioned, this allows anybody now to write their own archie client, something I would very much like to encourage. Who knows, we might have a nice X interface to the system sometime soon :-). This has many advantages: (1) The ability to have customized clients to provide the user with the functionality s/he desires. At least 80% of comments we get about archie now concern changing the format of the output or providing minor enhancements to the user interface. The simplest way to go about this is to provide a server and a protocol which gives you all the info you'll ever need. It's then up to the client programmer to filter and massage this information in whatever way they want. (2) The concept of archie "domains" whereby the global Internet would be split and an archie server designated as "primary" for its region. This simplifies the administrative procedure of keeping track of 680+ (and rising daily) sites worldwide. Although this concept originally had every archie site keeping a listing for every archive site, I've been thinking recently of just maintaining those sites for which the archie server is primary contact. This would fit into the Prospero and WAIS protocols very nicely, since both have the concept of "merging" different data sources. The wider you spread your search net, the more information you'd get back. However it causes problems in other areas. Possibly only field testing will resolve this. (3) Not the least problem at the moment is the load that archie puts on the host machine. Primarily for a lack of time, we have not been able to incorporate new searching routines which promise to significantly speed up access times and thus lower total load. Bill is currently working on this and we hope to have them online "real soon now". quiche is now averaging 1300 logins/day (120,000 interactive logins since Dec 90) and is often saturated. (4) There is an archie server in production in Australia and Europe seems to be in beta testing. There have been problems bringing archie online because of certain implementation flaws, primarily memory usage for database updates. Mea culpa. Also, we have been talking with a number of North American sites about providing archie service and hope to have them up soon. One of the big problems here is maintenance since there is no way currently to automate resource discovery (specifically anonymous FTP sites) on the Internet. (5) Some time will have to be found to incorporate WAIS into the current setup. I recently talked to Brewster Kahle from Thinking Machines about doing this and things are starting to move along those lines. Clifford has suggested merging the server to incorporate both Prospero and WAIS. As usual, comments and suggestions are welcome -Alan