[Grok-dev] a nice research project

Martin Aspeli optilude+lists at gmail.com
Sun Feb 8 01:07:50 EST 2009


Sebastian Ware wrote:
> "If it does a lot of work [...] you might need to understand a lot if  
> it breaks [...]"
> 
> That was the worst crap I have seen in years. Seriously, it's like  
> listening to someone advocating intelligent design over darwinism.  

Is it? I've pdb'd my way through the Zope 2 ZPublisher and security 
machinery more times than I'd have liked to, and I can safely say that 
if these had been shallower (and better commented), I would've been a 
lot happier.

I haven't done the same to Grok, but it's probably not a bad metric to 
say that if the call stack for a simple request is very deep, then 
there's at least some scope for improvement. That was Martijn's point 
when he started this thread. I don't think he wanted anyone to get into 
a religious debate. I also don't think Grok really needs people to 
defend it quite so vigorously. It comes across rather insecure, if not a 
little immature. I have only used Grok briefly, but I am pretty sure 
it's got enough going for it to be able to stand up on its own merits, 
with factual comparisons, rather than hyperbole. This is really the 
worst crap you've heard in years?

> I've been involved in scalability testing on the .net platform and in  
> my opinion this is just amateurish at best, ignorant at worst. 

Chris wasn't testing scalability. He was making a point about framework 
complexity. He was also clearly trying to show that repoze.bfg scores 
well in this regard. That's his prerogative, and frankly, it's informative.

> That  
> said, all angels of profiling are interesting and beneficial to follow  
> up, but the blog post insinuates that (with some feable reservations)  
> a user can draw interesting conclusions from the comparison. I think  
> not. It is only useful for the framework developers and the post  
> should be targeted at that audience.

I think you may want to read the post again. I don't think Chris said 
anything that was particularly contentious.

I suspect the reason why you're up in arms is that the Grok call stack 
is about 7 or 8 times worse than the nearest competitor, which does 
indeed make it look a little bad. I can't find a value judgement about 
Grok (or any other framework) in the whole post, though.

I suspect the long call stack has something to do with Chris' setup, or 
possibly something to do with the default Grok configuration. Martijn 
started this thread to find out why, and whether anything can be done. I 
wouldn't be surprised if there were one or two small things that could 
be changed to take this from the 800's to the 200's.

> I think the knew catch phrase should be "Lies, damn lies and blog  
> posts" :(

That's just childish...

Martin



More information about the Grok-dev mailing list