Forum OpenACS Q&A: Re: opencs.org stops working at night?

Collapse
Posted by Michael Bluett on
There appears to be something wrong with the site search, results mostly don't include titles and the brief snippets. E.g. Searching for "Google" brings back results like: "Untitled https://openacs.org/forums/message-view?message_id=53769". Perhaps this is due to Google's visit and the subsequent load on the machine?

The cvs repository suggests that there isn't a robots.txt for OpenACS.org in CVS. I plan to exclude Google from attempting to post messages and log-on on my site using robots.txt (by barring robots from /forum/message-post and /register). I'm sure a few of us would be interested in any subsequent alterations to the robots.txt to exclude various dynamic parts of the site from Google's gaze.

For those unaware, Google is doing a "deep crawl".

Collapse
Posted by Tilmann Singer on
Dave Bauer and I have recently made some changes to the search on openacs.org, with the final goal to only index full threads instead of single messages, to avoid the cluttering of search results. It seems to work already for new threads - check this search:

https://openacs.org/search/search?q=powerbook+browse+compile+debug

note that the word 'debug' only occurs in the last message - previously search would not have found anything when the search spans the full thread.

What still needs to be done is to reindex all the old messages. I don't know why the abstracts are currently wrong, but after the reindexing that will be repaired.

Collapse
Posted by Andrew Piskorski on
Eventually, listing the highest-hit individual posts as part of the thread hit listing would be ideal, but even without that, just returning the thread hit in search looks like a big improvement. Thank you, Tilmann and Dave!

Hm, in Forums, although each individual post always has its own message_id, Forums gives a link to display just that one post (and its direct replies) only if the poster assigned a different subject line, rather than accepting the default subject. That's kind of disconcerting. Occasionally you want to give a link to an individual post, so there should be a (small, unobtrusive) link on each post for displaying just that post.

It'd also be very nice to have <a name="message_id_foo"></a> anchors embedded in the page too, so you could display the entire thread but zooming in on a particular post. The UI decision of whether to make display using those name tags the default behavior for viewing a single post could (and should) be postponed to a later date, just sticking the name tags into the Forums HTML hurts nothing and would definitely be worthwhile.

Simple Matters of Programming, I guess. I'm sure others have thought of this too, just figured I'd point it out anyway...