Re: Subversion

On Jul 1, 2005, at 9:33 PM, Jeremy Kemper wrote:

>> I don't really see how that is applicable here. 1) we have a much  
>> much smaller repo, even after duplicating out all of the branches;  
>> 2) ruby has a very sane branch methodology; 3) I worked at a  
>> company that had the largest integrate DB (merge metadata) that  
>> perforce has ever seen (gawkable in size, it was 6gig alone), and  
>> yet the server did fine unless a developer did something truly  
>> horrific. This was in a 5m LoC source base tho with 50-100  
>> branches. I sat next to the p4 admin. Sure, they made mistakes  
>> that caused pain now and then, but most of that is avoidable by #2  
>> above.
>>
>> I highly doubt we'll ever have a problem.
>
> This is orthogonal: the problem is that the size of the depot  
> scales with the number of users accessing it.  The number of files  
> in source control is no issue, p4 can handle any project capably.   
> But each clientspec is ~20MB in size, so you are forced to limit  
> access or buy a big box *due to the number of clients*, not the  
> size or complexity of the repository.

20megs for a clientspec? Where are you getting these numbers and how  
would they actually relate towards ruby? If anything, the increase of  
metadata for a new synced up client is proportional to the number of  
files included in that client's view. That just makes sense. So, if I  
am to believe your 20meg assertion is true (and currently, I simply  
don't), it'd have to be for a rather large checkout. In my own  
experiments, I get an increase of 122kb checking out approx 3k files  
into a new client from my own repository. Ruby has 119 files.  
Assuming that these numbers are all proportional, that means a client  
with one version of ruby would take up approximately 4kb in the  
db.have file.

So, again, I doubt we'll ever have a problem with this if we decided  
to take on perforce.

> Please, I think Perforce is a good product, but it does have its  
> limits and large-scale public access is one of them.  Please pardon  
> me for contributing to this errant thread; advocacy is hardly  
> germane on -core :)

Yes, it does have limitations. I don't think that any of them are  
insurmountable and would like to see perforce evaluated fairly with  
objective numbers that come from cited sources.

--
[e-mail] - Seattle.rb - http://www.zenspider.com/ 
seattle.rb
http://blog.zenspider.com/ - http://rubyforge.org/projects/ruby2c