Re: Client-side searching proposal

Daniel W. Connolly (
Fri, 27 Jan 1995 20:13:50 +0100

In message <ab4ebb0809021004315d@[]>, Nick Arnett writes:
>Good point. When you get down to it, highlights could be considered to be
>a search engine's annotations! A general annotation scheme might thus be
>useful for many purposes. Interesting ideas would be gizmos that would
>link document sections to FAQs, as sort of an automated "explain this"

Ah... this touches a nerve. There are a _lot_ of applications of
this sort in action today:

* hypermail takes RFC822 messages and adds links, and reformats

Problem: what if I want the original RFC822 message? (why? because
it's digitally signed, or for other reasons of authenticity) I can't
get it through most hypermail archives today.

* Folks take a plain-text FAQ, or an interesting article, or whatever,
and they "htmlize" it. Sometimes they put a link to their home page,
or to the server's home page or whatever.

The point is: I would like to have some "audit trail" for such annotations.
I would like to be able to check the HTML version against the original, maybe,
or recover the original for any number of purposes.

It seems to me that the Right Thing To Do in they hypermail case, is for
the server to send the orignal RFC822 message, plus some annotations,
and then have the browser compose them -- or not!

But this is not optimal: you have zillions of browsers repeating the
work. Plus, the "composition" might be really involved and different
applications will want to do it differently, and it may require access
to a big database that only the server can see and...

One solution is to give the client the option of retracing the audit
trail with something like:

<link rel="original" href="original-msg.txt">

message with annotations...

More later... gotta go.