Re: urn-5 vs hashes (was: Re: [Gzz] RE: [ba-ohs-talk] Fenfire, RDF(re "Towards a Standard Graph-Based..."))
Hi Grant-- (01)
Grant Bowman wrote:
> * Benja Fallenstein <b.fallenstein@gmx.de> [030310 07:47]:
>>Kevin Keck wrote:
>>
>>>Actually, there's at least three used with the various file sharing apps:
>>>
>>>MAGNET, apparently supported by some Gnutella programs. Uses sha1. Example:
>>><magnet:?xt=urn:sha1:ZETNTXVNDO2DS32627BQNKIADS3RT3T5&dn=Shareaza1700.exe>
>>>
>>>eDonkey ed2k URIs, which use MD4. Example:
>>><ed2k://|file|Shareaza1700.exe|1330539|b29e2440d84ed468a63971220667786f|/>
>>>
>>>sig2dat URIs, used with Kazaa, Grokster, etc. Not sure what hash. Example:
>>><sig2dat://|File:%20Shareaza1700.exe|Length:%201330539%20Bytes,%201299KB|UUH
>>>ash/kHiGJMFBOMT2rdMU=|>
>>
>>Sorry, I should have been clearer, but by URI scheme I meant registered
>>URI scheme. (I believe in the power of standards :-) ) If I'm not
>>mistaken, none of the above is registered.
>
> Standard is a debatable word. What good is a standard that's
> architected perfectly but NEVER implemented and used? I would rather
> have a de-facto standard that is useful and helps people accomplish
> things on a daily basis over an academic standard any day. (02)
I need a bit of context here. Are you talking about a specific design
decision, or are you just making a general statement? (03)
I originally said there's no URI scheme for hashed data as far as I
know. As an end in itself, the issue of 'is there an URI scheme that...'
is a rather academic question :-) (04)
But for us there's a practical background: As a supporting technology
for Fenfire, we're developing a system called Storm in which all
information is stored in 'blocks' (immutable byte sequences) identified
by cryptographic hashes. One of the stated design goals is that, as long
as the hash function isn't broken, blocks stored in our system can still
be used in, say, thirty years-- provided there is still an
implementation of the then-current version of the spec around. Now, we
need a way to map these blocks into URI space. (05)
If we talk about how to identify resources for today or the next year,
there wouldn't be a problem with using an ad-hoc URI scheme. If we talk
about how to identify resources in 30 or 40 years, burdening future
developers with non-standard schemes (and possibly handling collisions,
should someone in the meantime decide to use e.g. sig2dat in a--
possibly only slightly-- different way). If I use only standard URIs, I
know that there is an official definition of its workings from the IETF,
and that the change control for it is at the IETF. (To draw an analogy
that's as wrong as any, look at how difficult HTML has become to parse
because of the many nonstandard extensions; I'm quite confident that
there will be *some* implementation of conformant XHTML 1.0 in thirty
years, but would you guarantee me that I will still be able to view my
Netscape 4-specific DHTML?) (06)
If there were any registered URI scheme meeting my needs, I could
implement it even if it were a purely academic standard so far. But I
would not want to commit to an ed2k or sig2dat URI for my data now. So
the question of whether there is a *registered* URI scheme or URN
namespace for this has very real consequences, for me. (07)
- Benja (08)