Filterset.G - I call "bullshit" · 2007-02-13 15:01 by Wladimir Palant
I stumbled upon the Filterset.G article in Wikipedia and noticed that G recently edited it adding some “facts”. The most ridiculous one is: “There are approximately 1.5 million users of Filterset.G”. And these statistics are supposed to back it up. From the look of it G simply took the absolutely meaningless number of sites per month.
But now that we can take a look at the stats we can get the real facts. Let’s take for granted that users updating Filterset.G manually are a tiny minority, most of them use the Filterset.G Updater extension. Looking at the URL hits in January, there were roughly 5 million downloads of the
filtersetg directory listing. This directory is downloaded by Filterset.G Updater on each update and the request always goes to pierceive.com — the actual filter list is downloaded from Coral CDN most of the time so we don’t have good stats on that. This extension will attempt an update every 2 days, lets say that successful updates are on average 3 days apart. That’s 10 downloads per month for each user giving us an estimate of 500.000 Filterset.G users (and this is likely to be an overestimate).
For comparison: Webalizer stats for EasyList show almost 4.5 million downloads in January. Given EasyList’s suggested update interval of 5 days this translates into over 700.000 users (actually more like 800.000 towards the end of the month). G, thanks for letting us know the real proportions.
Now lets go to G’s response to criticism (most of which originated here):
Some features specific to Adblock Plus that cannot be disabled create inconsistencies in behavior between Adblock and Adblock Plus with the same filters.
And he links to a discussion he had with me long time ago about whether Adblock Plus should be able to block stylesheets. I wonder what the “cannot be disabled” part refers to since last I checked Filterset.G was still disabling this feature with the “@@$stylesheet” exception rule. My point still stands: Adblock Plus should be able to block all requests and that includes stylesheets, without this feature it would be incomplete. It is especially very useful in Thunderbird — there I want to block everything by default to prevent spammers from using web bugs (that can be stylesheets as well as anything else) to track me. And the filter
/about.com/\d/(?!.*\.js)/ is still scary and is bound to have false positives. So the real problem is that Filterset.G is full of filters like this one.
Whitelist rules that inadvertently prevent ads from being blocked should be reported on the Filterset.G Forums, however the Adblock Filterset.G Updater keeps deleted filters from being downloaded on subsequent updates, so users can “fix” problems without waiting for an update to Filterset.G.
Great to know that Filterset.G Updater lets users “fix” the issue themselves (judging by the reports I get most of them didn’t discover this feature yet) but Adblock Plus has this functionality built-in. Which doesn’t solve the problem — in order to “fix” the issue users first need to understand what the issue is. And in most cases they simply say: “Adblock Plus is broken, it cannot block this banner”. So the real solution would be to cut down whitelisting rules to an absolute minimum and try to be as specific as possible with those. A false positive with a whitelisting rule is very bad. And Filterset.G is simply asking for them with rules like “@@;ad;” or “@@player;”. Ok, so any ad serving script that happens to have “;ad;” in its address is automatically whitelisted — how crazy is that?
The downtime issues documented above have necessitated the use of a separate Add-On in order to prevent (and rapidly respond to) further server problems.
How that? If G means that he can change the address in the extension — the “rapidity” of this response will be very limited. My experience shows that it takes at least 2 months until a new extension version reaches an acceptable level of adoption (maybe 90%). And I don’t see any code in the Filterset.G Updater to allow changing download URLs dynamically without updating the extension (this feature is planned for the next Adblock Plus release — and yes, it is long overdue, sorry).
On the other hand, Filterset.G not being served at a fixed location actually created the server problems cited above. EasyList’s 4.5 million downloads resulted in less than 5 million hits for the whole domain. Filterset.G’s 5 million downloads on the other hand became 17.8 million hits because every download requires 4 server requests where there should be only one (with Adblock supporting whitelisting there is no more excuse for keeping blacklist and whitelist separated). That’s what killed the server in the end.
Commenting is closed for this article.