Pony to get only new comments loaded on refresh?

February 27, 2002 4:09 PM

From the I Don't Know If This Is Possible Dept.: Could MeFi be coded so that when hitting the preview button on a new post, the next page only lists any posts made to the thread since the person last loaded the page? I'm just thinking in terms of lightening server load, since on some of those active, >100-post threads, the server's having to reload and resend the entire thread just so the poster can confirm his/her response looks right before hitting "post."
posted by aaron to Feature Requests at 4:09 PM (7 comments total)

There was already some discussion on that here.
posted by bingo at 7:26 PM on February 27, 2002

bingo, I don't see anything about cutting down the size of preview pages, there.

I think this is actually a pretty good idea, as it seems it could really cut down the load. It definitely is possible, and I don't think it would take much coding (though only Matt knows for sure). The only drawback might be that sometimes you want to refer to a previous post as you're writing your own (thus, during preview). Perhaps just showing the last [x] (5?, 10?) posts would be better.

I'm not sure how much of the load comes from commenting, though. I have a feeling it's a lot less than just plain old page views.
posted by whatnotever at 1:59 PM on February 28, 2002

whatnotever: Not exactly, but it seems to me that it's the same basic concept as what I was proposing. Which was that comments would only be listed as "new" if the individual user had not read the comments. Either way, you are making the system keep track of exactly which comments you have read and which ones you haven't read, as opposed to how recently they've been posted.
posted by bingo at 8:45 PM on February 28, 2002

I think this is different. In order to post, you are already in the thread, so it can be assumed you've read all the comments. The system can put a timestamp in the page when it is generated, then pass this to the preview page - any comments added since that timestamp are new.
posted by whatnotever at 7:49 AM on March 1, 2002

It still sounds to me like essentially the same mechanism. The timestamp has to be made either way.
posted by bingo at 9:05 AM on March 1, 2002

But in this case, the timestamp is just part of the served page, whereas the topic of the other thread involves writing a timestamp to a database every time a user reads a thread. Creating a timestamp is easy, and writing it into a page is easy, but writing it into a database is more complicated.
posted by whatnotever at 4:02 PM on March 1, 2002

Ok...well, sounds like you understand Perl better than I do, so I suppose you're right.
posted by bingo at 8:00 PM on March 1, 2002