Re: Smoothing the edges by example (was pilot case?)

On Wednesday, May 8, 2013 at 4:36 PM, François REMY wrote:

> > Documenting the experiences would be  
> > of great value to the community, we  
> > just need to do it (once we get enough  
> > experience).
> >  
> > I have experience from prollyfilling a  
> > few APIs with the WebIDL project… and I
> > plan to do a few more in the near future
> > (for SysApps).
> >  
> > I'm happy brand that work as having come
> > from this CG.
>  
>  
>  
> Nice. I agree that experience is a prerequirement on that front. Brian has a good and growing experience on prollyfilling CSS from which we can build upon, too.
Absolutely. The problem is that we have no criteria to measure if what we are doing is any good or not. Mechanical tools aside (linting or whatever), knowing if the quality of what we are doing is any good requires independent evaluation from our peers (who may or may not know what they are talking about). Best we can do right now is share our experience and pain points - and hopefully users of our tools can do the same.  

Alternative is to look at prollyfills that are used in the wild by large number of developers (e.g., parts of JQuery, picturefill, etc.) and see if we can spot some patterns (and come up with reasons as to why we think those are good patterns). And evaluate where the developer of the prollyfill has struggled because of platform limitations. If we find those limitations, we can then take them to various WGs and say: "see! people are trying to do X and they can't - we need primitive X". For example:
http://stackoverflow.com/questions/5136727/manually-artificially-throwing-a-domexception-with-javascript

We can also present these patterns to, for example, public-script-coord/TC39 and ask them if they are idiomatic uses of javascript.  
    
With WebIDL.js, we have basically 0 users right now (as it's not done and not yet practical to use in a project). I'm trying to grow that to a handful with specs I am prototyping, but so far it has only been me doing the prototyping.   

Brian can probably reveal his usage stats for Hitch.    

Hope that makes some sense…  

--  
Marcos Caceres

Received on Wednesday, 8 May 2013 16:05:40 UTC