Re: [Gnu-arch-users] Re: patch-log sizes

[Tupshin Harper]

Tom Lord wrote:

>    > From: Tupshin Harper <address@hidden>
>
>    > This reminds me of a question that I've been meaning to ask for some 
>    > time. Is there a strong reason why it makes sense for arch to store 
>    > files in compressed tars as opposed to storing compressed files in 
>    > uncompressed tars? Specifically, the difference between a .tgz 
>    > containing many .txt files vs a .tar containg many .txt.gz files. 
>    > (Extensions used just to illustrate the fundamental nature of the files 
>    > in question). Presumably compression ratios would be somewhat less, but 
>    > accessing a limited subset of the files in the tar would be *much* cheaper.
>
>    > Any thoughts? What am I missing?
>
> Revision libraries are a better way to provide cheap access to
> individual source file trees.
>
> Remember that most revisions are stored as changesets so that the .tgz
> is going to contain diffs, not individual source files.   Are you
> really sure you want to optimize access to those diffs?
>  
Absent any substantial costs, of course I would. ;-).

> (I would guess, (but it is just a guess), that the compression ratios 
> would be "much" less, not "somewhat" less.)
>
> -t
Yeah...on reflection, given the extremely small size of most of the 
content, you are probably correct. It seems like a bit of a pity, 
though, that there doesn't appear to be a good, portable, "free", 
compressed archive format with inexpensive random access.

(does web search and comes up with 
http://www.nongnu.org/duplicity/new_format.html as an example of 
somebody in the planning stages of addressing this need, though I 
disagree with some of his approach)

-Tupshin