[Forces-protocol] Re: Data encoding -- first part

[Zsolt Haraszti <[e-mail]>]

On Mon, 2004-10-18 at 16:58, Joel M. Halpern wrote:
> Why not just declare that structs start with their length.  Then we can 
> declare that an array consists of its element count (or length and element 
> count) followed by the representation of its contents.  [Ignoring for the 
> moment the issue of dense vs sparse packing, subscript representation, etc.]
> 

That's because when the struct is of fixed size, it is really not
needed and completely destroys even the possibility that you can
memcpy() a large array of these structs.

We may have put too much stress on the efficiency design goal.

But as we are talking to customers and partners, they consistent
concern is efficiency.

Therefore we decided that we try to design the data transfer such
that it scales 1:1 with the data size for large data elements, i.e.,
tables.  If your table is stored on X MBs in memory, it should ideally
require X MBs to transfer it over the wire.

/zsolt

> Yours,
> Joel
> 
> At 04:45 PM 10/18/2004 -0400, Zsolt Haraszti wrote:
> >This is another trade-off issue.  You are absolutely right, inserting
> >the size info in front of each struct can be very useful, especially if
> >it is a variable size struct.  We will have this in ARRAYs, not
> >part of the struct itself, but part of the ARRAY encoding.
> >That does not help when the STRUCT is embedded in another struct though.
-- 
Zsolt Haraszti                Phone:  [phone]
Modular Networks              Mobile:      [phone]
                              Email:  [e-mail]


_______________________________________________
Forces-protocol mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/forces-protocol