>It is quite useful to make this item a drop down menu so the user >can select the type of sort at search time.
My main concern with providing lots of search options is that they can detract from the human interface (I think the George Foreman grill, with zero buttons, is the pinnacle of industrial design). So the less intrusive to the majority of users who don't care about sorting control, the better. One way of doing this is making sure such control only appear on results pages, or under some "advanced" search option. Or if date sorting is obviously better, that should be the default and forget about giving searchers a choice. I'm willing to accept a patch for this, probably against src/wrapper.int (see FAQ for mail-archive source code, which is licensed under the GPL, not that anyone other than Mail-Archive runs it). Any such patch should be "tested" by doctoring up a Mail-Archive search form and submitting against the live htsearch on Mail-Archive to make sure it works. Localization of new words will be a pain, although I probably need to do a new round of that anyway with the "EMAIL PROTECTED" string. Finally, in anticipation of the perennial question "why not dump htdig?" for Lucene or Kir's excellent search engine aspseek. The reason, besides inertia on my part, is that Mail-Archive has now invested quite a few CPU-years in creating 150GB+ of htdig index files. So anything that requires re-indexing is pretty much out the window for the time being. >Sorting by date doesn't seem to work as all the messages seem to have >the same date, from I guess when it was last indexed ! Mail-Archive does create HTML message pages in batches, but the batch duration should not span more than a few days. Batch duration is directly related to archiving latency and is in the best of cases a few hours, worst of cases about a week. (And trust me, when archiving latency is a week, that's probably because something is wrong, and I hear about it from from people who follow lists via their archives). The filesystem timestamps for HTML files created on a batch run will probably all be the same, because the granularity of file timestamps only goes down to a minute. The htdig indexer reads straight from the filesystem, so I suspect it looks at filesystem timestamps. I would be very surprised if the time of search engine indexing had any effect at all. _______________________________________________ Gossip mailing list [EMAIL PROTECTED] http://www.mail-archive.com/cgi-bin/mailman/listinfo/gossip
