RE: Last version of the new summary storage engine

[<Dirk-Jan Binnema nokia com>]

Hi Philip, 

> -----Original Message-----
> From: tinymail-devel-list-bounces gnome org 

>   o. Duplicate strings are made unique in the mmap
> 
>      -> Current summary storage scatters duplicate strings 
> around in the
>         mapping. Making the mapping more random (less localised memory
>         access), more redundant (larger VmSize).
> 

Philip, I am a bit skeptical of this claim -- can you back that
up with some numbers? 

I mean let's take some *extreme* case, where the
same 10-byte string is seen 1000x in a mailbox; by having only
one of them we save about 10k. Not very much compared to the normal
size, even in this extreme case. At least the mem saving is
rather easy to calculate :)

What is much harder, there's also a (significant?) *cost* to scan for 
duplicate strings, to reconstruct them and so on, when showing a folder.

To me at least it's not obvious that this optimization will not actually

slow things down.

So, before rolling this out, I'd recommend you to do some numbers of the

difference with/without string clustering (this should include showing
the items).

Best wishes,
Dirk.