Subject | Re: Doc/spec for Gbak backup format |
---|---|
Author | Frederic G. MARAND |
Post date | 2004-05-31T07:15:23Z |
Helen, Ann,
Thanks for your suggestions. At least is provides a starting point :-)
If/when I get the format straight, I'll submit it to docs, then.
Frederic.
Thanks for your suggestions. At least is provides a starting point :-)
If/when I get the format straight, I'll submit it to docs, then.
Frederic.
> Open a gbak file in a text editor. There you can see what it'sfirebird-docs?
> storing. It starts with DDL for the system tables; then a bunch of DML to
> insert the object definitions; then a whole lot of alphabetti which I
> assume is the data in XDR format. That's as much as I can tell from a
> fairly cursory look. But, with the sources from BURP and a gbak file, you
> should be able to make some sense of it. That's all I can suggest.
>
> Oh, one more: when you do it, would you please submit it to
> :-))[...]
>
> /heLen
> From: "Ann W. Harrison" <aharrison@...>
> Subject: Re: Re: Doc/spec for Gbak backup format
>
> At 04:59 AM 5/30/2004, Frederic G. MARAND wrote:
> The difference between transportable and other in gbak, as
> in the remote client and server, is the use of XDR to create
> a machine independent format. The non-transportable format
> stores data in the backup file according to the data
> representation and alignment rules of the local system.
>
> The basic gbak file consists of a header, description of
> the global database characteristics, table definitions, then
> data. The format is generally tagged fields where the tag
> is defined in burp.h. In the case of variable length data,
> the format is "clumplet", which is a tag followed by a
> length, followed by that many bytes. In the file backup.e,
> look at the write_xxx functions to see how the data is acquired
> and represented.
>
> Borland never had documentation on the gbak format that
> I know of. There really isn't much design documentation
> except some written to describe a proposed feature. Since
> backup is ancient, it's protocol was considered self-documenting.
> And completely obvious to the designer...
>
>
> Cheers,
>
>
> Ann