[wp-hackers] Code reviews for plugins?

Mike Schinkel mikeschinkel at newclarity.net
Mon Aug 23 03:21:23 UTC 2010


A viable solution seems reasonably clear to me

1.) Crowd-source the code review with a mechanism similar to http://wordpress.stackexchange.com
2.) Focus on the top 15% of plugins by download and use
3.) Review based on objective criteria (i.e. adds tables and/or uses existing tables, use of nonces, list of symbols added to global namespace, number of wp_options added, etc.)

It need not be perfect, it just needs to give some objective and reasonable facts to allow people to decide their own guidance.

-Mike

On Aug 22, 2010, at 11:08 PM, Mark E wrote:

> 
> 
> On 08/22/2010 03:15 PM, Jeff Chandler wrote:
>> I wrote a post back in 2009 asking if something like what is bring
>> described here is nothing but a pipe dream.
>> 
>> http://www.wptavern.com/is-a-plugin-validation-team-a-pipe-dream
>> 
>> May be worth reading as well as the comments attached to the post. It's
>> a topic and suggestion that keeps coming up but no one seems to know how
>> to tackle without the intricate process of vetting one plugin against
>> another.
> 
> > What happens if a plugin has been reviewed and flagged as
>> awesome but a week later, has a security vulnerability discovered.
>> Doesn't that make the whole system broke and worthless?
> 
> The answer is simple: Ditch the review process, because obviously nobody involved is qualified to manage it.
> 
> 
> That said, there are issues about code review that people will not agree on - nor should they. Examples:
> 
> Some people are entirely anal about insisting on the use of camel case.
> 
> Other people are hell bent on using a class even if it only wraps one or two incredibly basic functions.
> 
> Other people insist that all your css goes in a subdir, all your forms go in a subdir, all your JS goes in a dir, etc - even if your entire plugin is entirely user-facing and can be contained in one really really really small file.
> 
> What all that boils down to is people saying "you need to write code like ME - I cannot tolerate diversity"
> 
> That sort of thinking has absolutely no bearing in a code review process unless code might be intended as a core patch, or core addition - that's different since that code ought to integrate seemlessly into existing style.
> 
> Personally, I don't care how people write code as long as three things happen:
> 
> - No potential function or class name collisions (e.g. be original and think ahead)
> 
> - No unnecessary overhead. If code isn't needed yet, don't load it. If you intend to use data, cache it internally.
> 
> - Think like an intruder and defend your code against becoming an vehicle for intruders.
> 
> So, in terms of code view for plugins, that's all I'd be looking for. All the rest is rather pointless.
> 
> That latter security item is the biggest hurdle.
> 
> One really really really big problem with policing anything is that as soon as you start you accept liability. If you fail at any time, you are perceived as being liable. On the other hand, if you don't police, no one can accuse you of doing a bad job and drag you into court over it.
> 
> That's the situation ISPs have faced for about 15 years. And so they don't police your activity or your content. They might spy on you, and they might reprimand you if someone brings your activity to their attention, but they don't police you. And thus they are held liable for what you do. They're only a conduit - just like the WP plugin and theme repo.
> 
> My opinion, if it isn't clear already, is to forget about screening the repo. That's not the responsibility of Automattic.
> 
> Mark
> _______________________________________________
> wp-hackers mailing list
> wp-hackers at lists.automattic.com
> http://lists.automattic.com/mailman/listinfo/wp-hackers



More information about the wp-hackers mailing list