On Wed, 27 Mar 1996 16:25:40 -0800 Peter Rauch
<[log in to unmask]> said:
>Good. No explicit mention of Packages. Since a file might belong to
>several Packages, I'm not ready to equate Package to directory contents.
Packages are Phase 2 items. It's not clear that many people still want
them. Nowadays packages are distributed as a ZIP file or equivalent.
Packages are probably not very difficult to implement, but users do find
the $PACKAGE confusing, and I can't wait for all the unix users to ask
why it doesn't work when they type $PACKAGE like it says in the manuals
:-)
>I'm not sure what this means. Web-based menus/point and click/reply to
>prompts/links is great stuff, but it's not database stuff. Much of what
>a discussion list is --is text. So, the graphical stuff you refer to
>must mean the interface and not graphical information content.
People want to click on "Search the list archives", then they want a
series of menus and forms allowing them to formulate their search. Then
they want to click "Go!" and see the results as a series of links that
each show a message. Something like that. What they don't want to do is
type DATABASE SEARCH commands.
Obviously you can provide forms that somehow submit a database job to
LISTSERV, but that doesn't give you a reply on the web. Or you can use a
local OS-dependent channel to LISTSERV to submit the database request and
collect the results, then make links. That may not be fast due to
LISTSERV being busy doing something else, it's limited to one concurrent
search, but it's one way to do it. Another way is to have a search engine
that can be used both by LISTSERV and in a web context. This is better
but you either use separate indexes or have an interesting file locking
issue with, if you're not careful, a potential for deadlocks. There are a
number of hybrid approaches where LISTSERV keeps the indexes up to date
and the web search engines can work in R/O mode. The whole web thing is
going to be a big can of worms, of which we're going to get a foretaste
when we release the archive browsing interface and everyone complains
that it's not easy to install and why does my web server do this that way
or require this or that and why do I have to configure so many paths and
so on, can't there be a GUI that figures what my web server is, goes
inside it and does everything for me? :-)
>See my comments above. The "simpler" syntax you imply wouldn't buy much
>more than echoing messages to a web/hypermail server and letting people
>do their searching, such as it is, there (which is what some lists have
>been doing, as I mentioned earlier).
Actually, I was thinking in terms of something similar to the VMS SEARCH
command, with a simple syntax, some context and an item number. If you
want the whole posting, you order it from the item number. Boolean
searches are not that hard to implement, the difficulty is in using
indexing, which could be done later I suppose.
Eric
|