Talk:Device API: Difference between revisions

From wiki.zmanda.com
Jump to navigation Jump to search
No edit summary
(Answers to questions.)
 
Line 1: Line 1:
= [[User:Dustin|Dustin]]'s Notes =
= [[User:Dustin|Dustin]]'s Notes =
Notes on things that confused me; results are pulled from existing device implementations, possibly with errors:
Notes on things that confused me; results are pulled from existing device implementations, possibly with errors:
* The volume label is assumed to occupy file 0, or at least, the data portion of the device is assumed to start with file 1
: The volume label is assumed to occupy file 0, or at least, the data portion of the device is assumed to start with file 1 --[[User:Dustin|Dustin]] 00:18, 17 September 2006 (PDT)
* File headers are assumed to be a part of the file structure; data in files begins at block zero (the test script seeks there)
:: Although the fd-device assumes that the label is at file 0, that is not an assumption of the Device API generally. File 0 is undefined as far as the public API is concerned, data files are numbered starting from 1. [[User:Ian|Ian]] 13:48, 19 September 2006 (PDT)


I'll add these to the page once I'm sure I'm right --[[User:Dustin|Dustin]] 00:18, 17 September 2006 (PDT)
: File headers are assumed to be a part of the file structure; data in files begins at block zero (the test script seeks there)  --[[User:Dustin|Dustin]] 00:18, 17 September 2006 (PDT)
:: Yes. File headers may be stored in the data stream (and extracted by the Device API), or they may be stored out-of-band. The difference depends on what makes the most sense for a device. For the S3 device, it might be best to store all the file headers in a separate metadata entirely, since amrestore might call device_seek_file() many times without calling device_seek_block() at all. [[User:Ian|Ian]] 13:48, 19 September 2006 (PDT)

Latest revision as of 20:48, 19 September 2006

Dustin's Notes

Notes on things that confused me; results are pulled from existing device implementations, possibly with errors:

The volume label is assumed to occupy file 0, or at least, the data portion of the device is assumed to start with file 1 --Dustin 00:18, 17 September 2006 (PDT)
Although the fd-device assumes that the label is at file 0, that is not an assumption of the Device API generally. File 0 is undefined as far as the public API is concerned, data files are numbered starting from 1. Ian 13:48, 19 September 2006 (PDT)
File headers are assumed to be a part of the file structure; data in files begins at block zero (the test script seeks there) --Dustin 00:18, 17 September 2006 (PDT)
Yes. File headers may be stored in the data stream (and extracted by the Device API), or they may be stored out-of-band. The difference depends on what makes the most sense for a device. For the S3 device, it might be best to store all the file headers in a separate metadata entirely, since amrestore might call device_seek_file() many times without calling device_seek_block() at all. Ian 13:48, 19 September 2006 (PDT)