On "Why mobile web apps are slow", garbage collection, and viral misinformation

A certain article has been enjoying great publicity and appreciation lately. I found some claims it makes to be unfitting to my existing knowledge of garbage collection (not claiming it to be much) so I tried to find out who was in the wrong and looked into the topic at some depth. Comes out that what we have at hand contains some nasty misinformation. Viral misinformation.

The following e-mails are an exchange between me and the author of "Why mobile web apps are slow", Drew Crawford. The article is to be found here, and from many other blogs linking to and praising it. The discussion is mainly centered around garbage collection, proving Crawford's claims to be based on factual inaccuracies, and showing his general lack of in-depth knowledge on the topic. This might seem like resorting to ad hominem, but I'd like to make a point of it since the tone and style of Crawford's article implies that he has a certain authority over the topic, which probably makes many readers take it as truth without an in-depth look into the factual accuracy of the presented information, or the coherence of the final interpretation of represented facts.

  1. forth
  2. back
  3. forth
  4. back
  5. forth

(There has been no reply after the last e-mail listed above, September 9.)

(Note that I'm publishing these e-mails as is (I fixed one typo of mine), and had not planned to do so prior; any significant mistakes of my own are preserved just as well as Crawford's. Specifically, my arguments on anything outside the topic of garbage collection are not particularly strong, I retract any hard claims there.)

Summary of the e-mail exchange

Other sources

Please also see this article, which touches some good points and seems, to me, to be written by a significantly more professional person (Shai Almog). One thing I dislike about the article is that despite how much it explains Crawford to be wrong, it softens up and claims that Crawford's article is "well written and very well researched" and that Almog "[accepts] almost everything [Crawford] said but [has] a slightly different interpretation on some of the points." This is possibly because the author didn't have time and/or motivation for controversy. :-)

There is also this benchmarking of the modern JVM GC, whose results show negligible performance difference at a mere 1.2x memory consumption. The page also mentions that the GC algorithms mentioned in the paper cited by Crawford do indeed seem to be typical primitives used to implement more sophisticated, "real world" collectors.

Hans Boehm, author of the Boehm-Demers-Weiser garbage collector, has a brief presentation on Memory Allocation Myths and Half−Truths.

Grand summary

At least on garbage collection, and very likely other topics touched in the article too (though I don't have proof regarding those), Drew Crawford unfortunately proves to be ignorant, and unaware and unwilling to accept of his ignorance; the article's claims on garbage collection seem to be the result of this ignorance coupled with self-misinformation and over-confidence. I apologize for the stark language.

The article cherry-picks an academic paper, which seems to have flaws of its own, then vastly misinterprets and misrepresents its results, making ridiculous claims. Crawford makes some attempts at justifying the content, but ends up proving further ignorance and a tendency to misinterpret academic papers.

I ask for anyone to not trust Crawford's articles without careful insight.

More importantly, I ask for everyone to remain skeptic on whatever they read, if it doesn't come from an acknowledged domain expert (and ideally even then), no matter how authentic the writing seems to be. These kinds of articles, looking great on the surface, yet containing huge amounts of well-presented misinformation, seem extremely dangerous to me.

Site design inspired by motherfuckingwebsite.com.

Feel free to copy anything from this page (including my e-mails), in part or in whole. I shamelessly published Crawford's e-mails alongside mine without prior notice, so perhaps refrain from doing the same.