Re: A Survey of ILP Account Identifiers?

On 11 December 2016 at 21:50, David Nicol <davidnicol@gmail.com> wrote:

> On Sun, Dec 11, 2016 at 2:51 AM, Melvin Carvalho
> <melvincarvalho@gmail.com> wrote:
> > On 11 December 2016 at 06:53, David Nicol <davidnicol@gmail.com> wrote:
> >> > In what context do you want to use SHA1 hashes?
> >>
> >> Mainline DHT cloud storage.
> >>
> >> > Is there a path to interoperability?
> >>
> >> Thanks to the interledger protocol project, ABSOLUTELY.
> >
> >
> > Could you elaborate?
>
> Uh, I'd rather finish my prototype first
>
>
> > So if I can understand the DHT cloud storage you are working on denotes
> > users with a sha1 hash (or is it files too?)
>
> It's everything -- everything gets a 160-bit handle to itself, as if
> all nodes are coexisting with a shared memory.
>
>
> > PS git uses SHA-1 too for files
> Yes of course; SHA1 is adequate for all practical purposes: the
> collision risk is small enough.
>
>
> > I am also working on distributed cloud storage but using http URIs as it
> has
> > the advantage of a large network effect.
>
> As does Mainline DHT.
>
>
> > Which systems have a path to interoperability, the cloud storage you
> > mention, the web based one I mention and ILP -- how does ILP leverage
> SHA-1?
>
> ILP, as a specification, needn't mandate implementation-specific
> details. As long as its agnostic to identity strings, using the
> 20-byte hash in an implementation would be equivalent to referring to
> an object by a memory pointer, while using the string would be
> equivalent to passing objects in some kind of marshalled form. From
> outside of the implementation, there should be no difference.
>

Thanks for the answers, I understand better now.

Im not sure you can put a 20 byte hash in the ILP ledger, I thought it was
just URIs, or does it allow plain strings of characters, too?

A sha1 is a random string of characters.

How is anyone going to tell that the string is a sha1 without some out of
band information?