Re: includehosts mandatory?

On Thu Jun  5 11:14:07 2008 Phil Archer said:

> PROPOSED RESOLUTION: That the <includehosts> element be mandatory for 
> all IRI set definitions.
> In favour 1: It gives us a way to ensure syntactically that an IRI set 
> is never empty

not quite true, as stray descriptorsets are, effectively, DRs with an
empty iriset. it, in fact, only restricts the number if syntactic
constructs that define empty IRI sets.

> In favour 2: It seems to feel right and generally make sense for our use 
> cases.
> Against 1: It places a limit on flexibility that may be unwarranted or 
> undesirable. Although not formalised (thankfully), a lot of web sites do 
> things the same way such as /images, /contact, /about etc. It wouldn't 
> be too hard to come up with a reason therefore one day to produce a DR 
> that described all resources on all domains where the path starts with 
> /images for example.

that's possible even with the restriction in place: there is a finite
and closed set of top-level domains, so one can easilly write an
includehosts that enumerates all TLDs, thus syntactically conforming
with the restriction, but semantically allowing all hosts.

> Against 2: <includeiripattern>, the WAF-inspired element, always 
> includes a host so you always end up with redundant elements if you use 
> that. Likewise if you use <includeregex> you don't necessarily, (but 
> might) need <includehosts>.
> I can't decide whether I'm for or against. I think I'm 55-45 against but 
> remain to be convinced one way or the other.

I suggest using to resolve this. Person A
matches odd and even numbers against a decision, and emails Person B.
Person C selects a file and emails Person B. Person B resolves the issue
by checking whether the first byte of the selected file is odd or even
(unsigned interpretation, zero is even). Person B announces the result
as well as A and B's choices, who confirm that these indeed were their


Received on Friday, 6 June 2008 04:03:36 UTC