Re: [Gzz] Questions [resent]

[B. Fallenstein]

Tuomas Lukka wrote:
> Benja: what about merge and including other cells? Just create with
> that ID, no transclusion etc?

Oh, important issue. Here's something to discuss.

Yes, just create with that id; no, no transclusion. BUT, there is a 
problem currently: Information about new cells is NOT SAVED.

With all the turns and changes in the overall scheme of things that have 
occured in the past year (most recently, SecureRandom cell ids), I now 
think I have good reasons to propose a scheme in which cell existance is 
wholly handled by structural means.

This means: All cells exist in the sense that Cell objects and 
connections can be created for them.

Additionally, there is a rank starting at the slice's home cell, along 
d.cells; each cell on this rank is considered to "exist" (be here) each 
cell not on this rank is considered not to "exist."

The distinction is important: Cells which do exist have been created by 
a user (or some program a user's called), and thus can be expected to 
have reasonable content/connections that somehow represent them. For 
example, a dimension cell may contain a text ("d.bazz"), or it may be 
connected to some other cells that cause an image to be shown inside the 
first cell.

Cells which have not been created will not have content/connections that 
make them showable nicely. An immediate consequence could be to show as 
their content their global ID-- maybe in gray and italics, showing 
clearly this is not normal cell content.

The point is that these cells can still have connections that allow us 
to say something about them. For example, we can implement transclusion 
structurally: We connect the transcluded cell to its identity along some 
dimension. The original cell doesn't have to exist in the space where 
the transcluded cell exists. Or, we can connect information about a 
dimension to the cell whose ID is that dimension's ID-- even if that 
cell doesn't exist at all in our space.

(Users, of course, should then get an interface for making a 
non-existing cell exist. This will be easy.)

Of course, this suggests an N(String) method in the Space interface 
(making that cell exist), which then Merge etc. could use-- I think 
that's reasonable.

More benefits of this scheme will likely become clear when-- at some 
distant point past 0.8-- we decide to implement the second stage of 
Ted's preflet scheme, with remote clones and so on. I think ultimately 
we won't have one cell per id and slice, but commonly one cell per id-- 
only if there are conflicts in merging different slice's cells together, 
there'll be two cells with the same global id (but different slice 
prefixes). But that's for the future.


Anyway, enough babbling. This is how I propose to implement cell status 
(existance) saving.

- Benja