[OSM-dev] Binary OSM; the first pass encoder

[Stefan de Konink]

Chris Browet wrote:
> 2008/11/9 Stefan de Konink <stefan at konink.de <mailto:stefan at konink.de>>
> 
>     Chris Browet wrote:
> 
>         Right you are.
>         If I understand the concept, that would mean that once a string
>         appear in the "master string" file(s), it's indice cannot ever
>         change (otherwise, all the references will have to be changed,
>         too). Do I understand correctly?
> 
> 
>     Yup. But as you might realise; it is good to recreate the files over
>     lets say months. Because some tags will not be present anymore; or
>     words contain typo's... etc.
> 
> Sure, but that would mean re-updating the whole database to update the 
> indices, same for local copies.
> That a bit denies the "diff" concept, don't you think?
> 
> Or you can leave "blanks", to be reused...

You are now trying to solve the fragmentation issue in every filesystem 
;) I think that blanking out the values with \0 could be solution for 
it. And allow on local side to 'defrag' the file so that LEN \0\0\0\0 ->  0.


Stefan