Re: Evolution of the RWW -- a Temporal Web -- Towards Web 4.0 (?)

On Mon, May 17, 2021 at 8:52 PM Henry Story <henry.story@bblfish.net> wrote:

> > On 17. May 2021, at 20:53, Melvin Carvalho <melvincarvalho@gmail.com>
> wrote:
> > On Mon, 17 May 2021 at 19:50, Henry Story <henry.story@bblfish.net>
> wrote:
> > > On 17. May 2021, at 03:33, Nathan Rixham <nathan@webr3.org> wrote:
> > >
> > >
> > > A loosely coupled immutable timestamped record of state changes allows
> deployment of resources to be broadly tech and protocol agnostic. For
> example several previous states of a document could be stored on IPFS, with
> the stateless protocol HTTP providing the most recent state, and a chain
> exposing timestamped pointers to the previous states.
> >
> > You can also get quite far just by adding link between version states
> using memento for example.
> > I am trying that out in the new Solid Server I am putting together.
> There are comments in the code here:
> >
> https://github.com/co-operating-systems/Reactive-SoLiD/blob/master/src/main/scala/run/cosy/ldp/fs/Resource.scala#L223
> >
> > Thanks Henry
> >
> > This looks interesting, I've had a look at the comment you pointed to,
> and it looks interesting, tho I didnt understand it fully
> >
> > Would it be possible to describe in a couple of sentences for those that
> are not intimately familiar with memento
>
> I have not implemented memento fully myself. It builds on rfc5829 and
> allows one to link to archives that keep
> copies of versions, though the server can be it’s own archive I believe.
>
> >
> > I can imagine a work stream for a temporal web based on this approach,
> if there's interest
> >
> > The thing that particularly interests me is what I'll term quite
> vaguely, a web-scale temporal web.  What I mean by this is, that the
> timestamp operation (aka the witness operation) is global scope and not
> local.  Meaning if any one website goes down, the timestamping record will
> still be there allowing a reconstruction of the history.  Providing
> resilience.  In 2021, we have specialized time stamping servers (commonly
> referred to as pubic block chain) which can provide this time travel type
> functions.  From your comment there is a time travel in memento too.  In
> any case, interested in your findings, if you'd like to share ...
>
> I think there is a clear need for any read-write web server such as Solid
> to keep versioning information,
> just to help restore versions in case a buggy hyper-App writes some data.
> That use case does not require
> global consensus on version states. So I think that is local versioning
> will clearly be the first
> priority. Trellis-LDP implements somehting like this too, so I am
> following in the footsteps to get a
> better understanding of it.
>

Agree, that's the trouble with using a stateless protocol to manage state!

Exposing version via headers is a neat solution, I guess you can implement
etag and if none match on write requests to avoid updating a resource for
which you have a stale state / which has already changed without you
knowing. That is probably good enough for most cases.

I guess chains come in to factor when you need reliable timestamps, and a
provably tamper proof immutable chain of previous states.

It would be nice if various approaches worked together.