[OpenAFS-devel] Re: IPv6 support

[Simon Wilkinson]

On 14 Feb 2011, at 14:04, omalleys@msu.edu wrote:

> Quoting Andrew Deason <adeason@sinenomine.net>:
>
> Maybe some documentation in the code where the changes need to occur  
> would be useful? If it is, then it would also be useful to have a  
> standardized status attached to the
> comment so you can search for it, ie // ipv6 support needed, //ipv6  
> support done.

I think the problem is that such an audit of the entire code base is  
likely to be a considerable amount of work, and require a reasonable  
depth of understanding of the code. For example, it's far from clear  
to the untrained eye which structures are serialised to disk, or are  
part of a Ubik database, and which are purely held in memory for the  
life of a process.

I think the best approach is the one that Andrew, Love and I are  
advocating - start at the bottom of the stack, work your way up to the  
top. For Summer of Code purposes this predominantly means adding IPv6  
addressing to RX. Fortunately, the RX wire protocol doesn't include  
address information (that's handled below it, in UDP), so this is just  
API extension work.

> 64-bit native server binaries?
> I assume most of this is going through the code and checking types  
> and code clean up, fixing alignment errors, etc.

I was pretty sure we did 64bit native binaries already. We've  
certainly been fixing bug reports related to 64bit support as we  
receive them (most recently in supergroups), but these seem to be in  
relatively untrodden areas of the code. Do we really still have 64bit  
issues on some platforms?

> A test suite functions?
> There are actually quite a few real jobs I have seen posted for this  
> for people with experience lately. Given the nature of this project  
> it would make sense to be able to give some support for testing it.

Extensions to the test suite would be invaluable - Russ has a  
framework in place for running these, although that's predominantly  
geared to low level unit tests.

>> That said, I thought we were still unsure whether OpenAFS was doing  
>> GSoC
>> this year...?
>
> Google supposedly expanded their GSoC money.

There are a number of hurdles to jump before even applying to Google.  
We need mentors. Please don't assume that those people who mentored in  
previous years will do so this year - many of us are significantly  
busier this year than last, and mentoring is a significant time  
commitment. Without a reasonable pool of mentors, there's no way that  
we can make a credible application to Google.

Cheers,

Simon.