Re: Global tags [was: HTTP Protocol Extentions]

Hi Larry,

Larry Masinter wrote:
> There's no point in having multiple "URIs" as a way of doubling
> the space. Maybe you could have multiple entity tags for a given
> entity, though.
>
> However, I'm now leaning toward just allowing the 'content-id' 
> (the official MIME version, not the CDF version) to be used
> instead of an entity tag inside an IF-MATCH, etc.
>
> After all, content-ID is a globally unique identifier. If you
> want one that's guaranteed to be based on MD5, then make it
> cid:md5:NNNNNNNNNNNNN@w3.org, for example.

That sounds like an interesting solution. Here are some
of the relevant rules:

>From RFC 2111:

  content-id    =  url-addr-spec
  url-addr-spec =  addr-spec
  cid-url       =  "cid" ":" content-id

>From RFC 822:

  addr-spec     =  local-part "@" domain
  local-part    =  word *("." word)
  domain        =  sub-domain *("." sub-domain)

My only concern about this approach is that a domain name is 
required in the content identifier. I'm not sure what domain 
name would be used for an MD5 checksum. Are you suggesting that
we pick a domain name which must be used for all MD5 checksums,
or are you suggesting that each vendor appends their own domain
name. That last solution would eliminate sharing between multiple
vendors.

One obvious thing to do is to make the @domain component of the 
content identifier optional since the MD5 checksum itself would 
be globally unique by itself. That would require a change to
RFC 822 (probably not a good idea). 

Have fun,

	Arthur van Hoff