Running a local copy on Validator.nu (was: Re: DogFood)

On Jan 14, 2008, at 11:41, Henri Sivonen wrote:
>
> On Jan 14, 2008, at 05:48, Sam Ruby wrote:
>
>> Also, I would be interested in installing the validator locally.   
>> While that appears to be possible, what I really would like to do  
>> is to keep my local installation up to date.  Thoughts and pointers  
>> on how to do that would be appreciated.
>
> The build script downloads the source from Subversion, downloads  
> dependency jars from around the Web, builds the source and runs the  
> validator. Running the script twice in a row should always give you  
> the latest version.
>
> There are three issues with keeping an installation up-to-date:
> 1) Self-updating of the script isn't polished. Hence, when the build  
> script changes, you need to run it twice: the first time updates the  
> build script and the second time runs the new build script.

I have fixed this. Works on Linux and Mac OS X. I have assumed Unix  
os.execv() semantics. I have no idea what happens on Windows.

> 2) There's no automatic notification mechanism for telling an  
> installation to self-update. Instead, you need to stop the instance  
> manually and rerun the build script when you know you want to sync  
> with the svn repo.

This one I haven't fixed. Do you have suggestions on a notification  
mechanism that would work for you?

> 3) The validator scrapes the spec for UI strings and, therefore,  
> needs a copy of the spec that is compatible with the scraper. The  
> spec isn't in the validator svn, though, so you need to give the  
> build script a file URI or an HTTP URI to a suitable copy. The live  
> spec isn't suitable, because it may become incompatible with the  
> scraper and the schema without warning. Right now, I'm hosting  
> snapshots of the spec and the About page says which magic URI is in  
> sync, but this again lacks a notification mechanism and is generally  
> bad. It seems like a bad idea to keep a copy of the large spec file  
> in the validator repo. But, then, perhaps keeping a copy of the spec  
> there would be the most practical thing to do after all.


This one I have fixed.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/