February 28th, 2008

The biggest mistake I ever made was to install Wordpress at the top level. The second, was to use "smart" URLs.

My site was restricted due to bandwidth overlimit this morning, something that shouldn't have happened. When I checked my stats, one site, proxyit.com, was hammering my bandwidth. Checking the recent visitors list, this domain was grabbing my feed every minute, except it was grabbing the Burningbird feed, which was then redirected to the new combined feed, at http://burningbird.net/feeds/atom.xml.

This feed, created by the aggregator, Venus, hadn't changed, but with the redirect, it was coming up fresh and sparkly new. Now, that doesn't excuse the fact that this site was accessing it every minute, but I'm not sure that my twisted convoluted redirects to feeds wasn't at least partially responsible. To make matters worse, I used an inline SVG object yesterday, which shouldn't tax bandwidth limits overmuch…unless your feed is being hammered.

(Not to mention that using SVG inline absolutely killed my entry at Planet RDF…)

Of course, when I redirected my Burningbird main feed, /feed/ to atom.xml, this redirected all other variations of /feed, including /feed/atom, /feed/rdf, and so on. Not just for Burningbird, but all sub-domains, too. So I had to add more redirects, which attempted to bypass Wordpress's programmatic management of URLs. I had to so many redirects in my sites to get feeds to serve correctly, I wasn't sure who was getting what. So I've removed all of them.

One of my site changes is to remove Wordpress at the top level. I'm replacing it with a page generated by Venus that combines all feeds from Wordpress installations in sub-domains. Each sub-domain gets its own Wordpress installation. Some will get the full installation, and others will get my new semi-forked version that I've named Curmudgeon Wordpress. Curmudgeon Wordpress is a Wordpress installation that has had all the reader interactive bits, such as ping back, registration, XMLRPC, and comments, and their associated includes and admin functions removed.

When I get all this finished, no more RDF feeds, no more RSS feeds. You get one feed for each Wordpress installation, an Atom feed. And you get one overall feed generated by Venus, name and location TBD, generated once per day.

In the meantime, feeds may be a problem. My bandwidth may be exceeded. Yada yada, you know the rest.

Comments
1
Seth Finkelstein - 6:03 pm February 28, 2008

By the way, I suspect that inline SVG causes some problems with browsers. The new page is extremely slow to redraw in my Linux/Firefox configuration.

2
Shelley - 9:46 pm February 28, 2008

It could also be my current hosting situation Seth. Things are not going as well as I thought they were. On the other hand, I don't test in Linux. The Firefox implementation of SVG could be a problem. Did you submit a bug?

3
Jeff Schiller - 10:47 pm February 28, 2008

Just as an additional data point: The Firefox 2.0 browser I use on Linux is quite a bit slower than its Windows counterpart. I haven't tested a recent Firefox 3 Beta…

4
Daniel O'Connor - 1:40 pm February 29, 2008

Using FF3beta3 / windows.

When I'm scrolling up and down, its a quite painful experience compared to a non inline svg page.

I think I can hear some component of my pc actually ticking under the strain…

5
6
Shelley - 2:36 pm February 29, 2008

I tried it myself with Firefox 3 beta 3 on the Mac and in Windows (XP), and I haven't had any problems with the inline SVG.

The fixed background image has always been a bit of a drag with Firefox and other browsers, but that's been that way forever.

I've also just optimized my images, and they're half the size. That could also impact on drawing, since I sample the image for part of the CSS. In fact, I'm going to be putting other optimizations in place, which should make the page load even faster.

Some of this could be due to the inline SVG, and some of the issues could be my (previously) larger image sizes, fixed background image, and other to-be-optimized playing around that I'm doing.

We'll see what happens with that bug. Thanks for turning it in, Daniel.

7
Shelley - 2:53 pm February 29, 2008

PS Jeff, can you let me know what happens when you test Firefox 3b3 in Linux?

I also did a CPU check with redrawing with Firefox in WindowsXP. I don't have a super system, but it is more than sufficient for streaming video, if we use this as a measure to go by.

To redraw my page by grabbing the scrollbar thumb and moving up and down, very fast, takes a maximum of 54%. Seth, yours ended up at 17%. Daniel, yours ended up at 19%. Jeff, yours peaked at 42%, and you're also using inline SVG. Sam Ruby, who uses inline SVG peaked at about 33%. Eric at ed.blog has inline and his uses about 22% max.

However, I changed my options to only display one post on the front page, which got rid of the Ajax-collapsed posts. That changed my CPU percentage from 54% all the way down to 22%.

I think we might find that the real draw is less the inline SVG, and more the use of JavaScript to collapse the older posts. So much so, that I'm thinking about dropping the use of JavaScript to collapse the posts (done).

PS I then closed the browser and tried Jeff and Sam's weblogs again. Jeff, the second run you peaked at 22%, and Sam peaked at 19%. Though not a very scientific approach, from what I'm able to see, I don't really there is an issue with inline SVG. I think there's an issue with Ajax collapsed content that is scrolled heavily.

Last PS And I removed the "fixed" for the background image and scrolling using the keyboard takes much less CPU and goes quite nicely. It's as I thought: a lot of the jerky motion with scrolling is probably because of the fixed background. Which I hated to remove, but until browsers deal with this better, I shouldn't use. Sigh.

In fact, once I removed the fixed background, my CPU usage is about equivalent to your's Seth. Between 18 and 22%.

So I would say the inline SVG is probably one of the better behaved options in my page. But all your own milage could vary. After I've made these changes, I'd be curious, Daniel and Seth, if you're still having the same problems.

8
Daniel O'Connor - 7:27 pm March 1, 2008

Ah, bugger. Shelly, any chance of putting up a static example page of what your blog *used* to look like?

… and commenting @ https://bugzilla.mozilla.org/show_bug.cgi?id=420343 ?

This was the URL in the bug, and now we can't reproduce the behavior…

9
Shelley - 7:32 pm March 1, 2008

Sorry Daniel, but Burningbird at http://burningbird.net still uses all the same functionality, and should still have the same problem. I'll update the comments to point to that weblog, and I'll leave it for a time.

Thanks to all those who have contributed to the discussion. Comments are now closed, but you can contact the author of the post directly.