[AFS3-std] DNS SRV Resource Records for AFS

[Jeffrey Hutzelman]

--On Monday, October 05, 2009 02:21:23 PM -0400 Jeffrey Altman 
<jaltman@secure-endpoints.com> wrote:

> I do not think that documenting an implementation deficiency is
> appropriate for a protocol standard.

I do.  It's entirely reasonable for a protocol spec to document common 
implementation pitfalls.  While it happens that current OpenAFS gets this 
right, it's actually very common for software depending on DNS data, 
especially name->address mappings obtained via gethostbyname(), 
getaddrinfo(), or similar interfaces, to fail to apply DNS-level TTL's to 
how long results are used at the application level.

-- Jeff