The Journal - a proposed syslog replacement [LWN.net]

The Journal - a proposed syslog replacement

Posted Nov 18, 2011 17:52 UTC (Fri) by mathstuf (subscriber, #69389)
In reply to: The Journal - a proposed syslog replacement by dlazaro
Parent article: The Journal - a proposed syslog replacement

Undocumented? If you're talking about systemd, it is one of the better documented projects. If your distro isn't announcing the change, that's not something you can blame on Lennart.

to post comments

The Journal - a proposed syslog replacement

Posted Nov 18, 2011 18:03 UTC (Fri) by jubal (subscriber, #67202) [Link] (5 responses)

No, the binary format that the journald will use won't be documented. Read the google doc that describes the design and FAQ. Also, journald. It simply had to be a generic name, hadn't it?

The Journal - a proposed syslog replacement

Posted Nov 18, 2011 18:25 UTC (Fri) by mathstuf (subscriber, #69389) [Link] (4 responses)

Ah, I got mixed up by the "are arriving" tense. I don't expect journald is anywhere other than their machines yet.

The undocumented format is annoying.

The Journal - a proposed syslog replacement

Posted Nov 18, 2011 18:38 UTC (Fri) by jake (editor, #205) [Link] (3 responses)

> The undocumented format is annoying.

My sense is that they are trying not to get trapped into backward compatibility games by having other programs that read/write the data without using the supplied library. I don't think they plan to deliberately obfuscate the format (and they will be providing code to both read and write it), but my guess is that they don't want to get stuck into a particular format forevermore because someone wrote a program that grabbed the 12th byte of every record and decided that some data would always appear there.

It's essentially the ABI problem that the kernel runs into, and that sometimes makes it difficult to change things in the kernel (like tracepoints for example).

jake

The Journal - a proposed syslog replacement

Posted Nov 18, 2011 18:47 UTC (Fri) by mathstuf (subscriber, #69389) [Link] (2 responses)

There still needs to be some backwards compatability mechanism because the tool will need to read logs in the older formats (at least, I sincerely hope so!). In that case, not documenting the format is silly. If the format isn't versioned, then the format is silly.

The Journal - a proposed syslog replacement

Posted Nov 18, 2011 21:29 UTC (Fri) by elanthis (guest, #6227) [Link] (1 responses)

The document states this is temporary. The feature isn't even in the main systemd code yet, but is on a separate branch. They're avoiding specifying a format until they're sure that it's nailed down, does what they need, doesn't have some unforseen problem that could only be found through widespread testing, etc.

Once it's a stable feature, expect the format to be documented. Until then, the documentation is available in the .c files for anyone who has some need to avoid the provided library interface.

The Journal - a proposed syslog replacement

Posted Nov 18, 2011 21:57 UTC (Fri) by mathstuf (subscriber, #69389) [Link]

> They're avoiding specifying a format until they're sure that it's nailed down, does what they need, doesn't have some unforseen problem that could only be found through widespread testing, etc.

This is understandable, even appreciated. Having multiple minor versions of varying usefulness indefinitely supported just because they existed for a commit or two during the inital format designs would be crazy to expect.

> Once it's a stable feature, expect the format to be documented.

FTA:
> At this point we have no intention to standardize the format. We might document the on-disk format eventually

This is not a strong a guarantee as I'd like for this and certainly not something that would range a high level of expectancy from me. Before it becomes default in a major distribution, I'd like to see a format specification.