Eric,
Allah be praised, we've struck a well!!!
Great news. Looking forward to reviewing the architecture, and seeing work
product. "Categories" can get very complicated, so your solution is a great
breakthrough.
Glad to see quiet engineering underway, as you noted on 000927, and that
engineering management is keeping up with the design so the architecture makes
sense.
Does this implement your CDS specs v0.9 on 000614?
Again, congratulations to you, and the entire team for this step forward.
Rod
Eric Armstrong wrote:
>
> Great summary of the issues I've been concerned about,
> Eugene.
>
> I'm happy to say that I have at last constructed some
> real, genuine source code that can form the basis for
> a distributed, versioned, categorized, link-typable,
> linked, node-graph repository. I've been developing
> the design document side by side with the code, so the
> architecture might even make sense by the time I'm done.
>
> Look for an announcement very, very soon now.
> (This week, I hope.)
>
> The code will have a lot of bones showing (large sections
> are still skeletal). And I'm sure there are some well-
> understood patterns I'm missing that could be applied to
> improve the solutions.
>
> In outline, though, I'm delighted with the design. One
> major nut to crack was how versioing works in a node
> graph. The solution to that came on the way to figuring
> out a solution the "competing versions" problem, which
> plagued me most of the weekend. It is pure heaven to
> think that a foundation might actually now exist that
> can be used to develop the functionality.
>
> Having defined a core architecture capable of supporting
> distributed, peer-to-peer operation, I'm a lot less
> concerned if the initial implementations use a simple
> client and put most of the burden on the server. I'm
> pretty sure that if that capability isn't built in at
> the outset, it will be harder than hell to retrofit
> later.
>
> Ensuring the basic functionality opens the door
> to a whole world of implementations -- one of which
> *should* be a browser-based interface, both for making
> it easy for others to take a look at it, and for
> convenience in accessing the system from some unclean
> system that is not enlightened enough to act as a
> peer-to-peer server.
This archive was generated by hypermail 2.0.0 : Tue Aug 21 2001 - 17:57:56 PDT