Comments on: GrabPERF: Search Index Weekly Results (Sep 12-18, 2005)

By: smp

smp — Tue, 20 Sep 2005 00:58:33 +0000

However, PubSub will still come up on top; Google doesn’t bother to compress their ATOM feeds.Ouch!smp

By: Bob Wyman

Bob Wyman — Mon, 19 Sep 2005 22:52:35 +0000

I'd like to suggest that you include a second test for Google BlogSearch to make the comparison to PubSub a bit more fair. What you're retrieving from PubSub is an Atom feed, not a full HTML page. Typically, HTML pages can be filled with all sorts of junk that isn't present in Atom feeds. Thus, the Google BlogSearch numbers may be getting dragged down by them having to transfer more data or do more formatting of the output. So, what I'd like to suggest is that you include a second "Atom" search test for Google BS that would look like this:http://blogsearch.google.com/blogsearch_feeds?s...>Since PubSub always returns 32 results (if present) it makes sense to adjust the Google BlogSearch item "num" to be 32. Admittedly, there would still be a difference between PubSub and Google BlogSearch since GoogleBS only returns summaries while PubSub returns full posts. However, this difference in payload size might get lost in the wash.bob wyman

By: smp

smp — Mon, 19 Sep 2005 19:58:33 +0000

Ok the test is added.

However, PubSub will still come up on top; Google doesn’t bother to compress their ATOM feeds.

Ouch!

smp

By: Bob Wyman

Bob Wyman — Mon, 19 Sep 2005 17:52:35 +0000

I’m a little surprised to see that the Google Blogsearch numbers aren’t better than they are. The site certainly feels much snappier than the numbers would indicate.

I’d like to suggest that you include a second test for Google BlogSearch to make the comparison to PubSub a bit more fair. What you’re retrieving from PubSub is an Atom feed, not a full HTML page. Typically, HTML pages can be filled with all sorts of junk that isn’t present in Atom feeds. Thus, the Google BlogSearch numbers may be getting dragged down by them having to transfer more data or do more formatting of the output. So, what I’d like to suggest is that you include a second “Atom” search test for Google BS that would look like this:

http://blogsearch.google.com/blogsearch_feeds?scoring=d&q=“new+york”&num=32&output=atom

Since PubSub always returns 32 results (if present) it makes sense to adjust the Google BlogSearch item “num” to be 32. Admittedly, there would still be a difference between PubSub and Google BlogSearch since GoogleBS only returns summaries while PubSub returns full posts. However, this difference in payload size might get lost in the wash.

bob wyman