billbill's comments on searching the archive

Date: Fri, 28 Jan 94 17:47 PST
From: billbill@wetware.com (BillBill)
To: bizarchive@media.mit.edu
Subject: more musings about what IT should do


Some things I want to see when I go into the archive (discounting
speed and coding complexity for now, paying close attention to them
later):

A search is specified by one or more of these terms, and may be
modified by use of at least AND OR and NOT operations:

Subject (of article)
Subject (of thread) (i.e. don't match articles after thread changed subjects)
Date (of article)
Date range (of articles)
Author (of article)
Author (of thread's/child-thread's initial article) (e.g. author of an
	article that is a branching point of a thread tree)
Length (of article)
Length (of thread)
Binaries (uuencoded images, others?)
Is not a followup (article)
Is a followup (article)
Has no "included" text (at least for ">" inclusions)
Is not a crosspost (article)
Is not a crosspost (thread) (I.e. thread contains no crosspost)
Is not a crosspost that includes newsgroups: ___[, ___]*
Is a crosspost (article)
Is a crosspost, including newsgroups: ___[, ___]*
Frequency of a given letter (t.b.scrabble)
(key)Word contained in (article or thread)
Randomness (P% chance applied to any hit criteria, or P% of articles
	matched are counted as matches)
Subject is unique (P% of words in subject don't appear in other subjects)

Raw search/grep/query string typed in by reader 

Also, there are some "NEXT" buttons somewhere on each article, to:
	keep the current thread and allow specification of a new one
	in the selected set, go to the next by this author
	in the selected set, go to the next in date+time order
	in the selected set, follow the thread branches from this article
	in the selected set, go to the next subject in alphabetical order
	in the selected set, go to the next in increasing/decreasing
	                     length order
	in the selected set, go to some random unread article
Maybe a simple specification of "the sort order to use right now"
along with a single "next in sort-order" button would work well.

There should be a table of existing author home-pages. Any reference
anywhere that includes the author's mail address is a reference to
this home page.

All article text is marked as <pre>-formatted text. Maybe if authors
send email saying "Article <message-ID> is ok to reformat," that
message can get tagged somehow and be displayed in nicer proportional
fonts. Authenticating authors would be a drag, though.

If I'm really browsing, I'd like to specify a search, then a new
search on the resulting articles, then a search on the resulting
articles... not just a single search specification. A single search
specification (the AND of all the searches I did) could be a result
document though, and I'd like to see it and save it, so I can quickly
and easily return to the set of articles I found.

Bill