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

The Journal - a proposed syslog replacement

Posted Nov 18, 2011 18:47 UTC (Fri) by mathstuf (subscriber, #69389)
In reply to: The Journal - a proposed syslog replacement by jake
Parent article: The Journal - a proposed syslog replacement

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.

(Log in to post comments)

The Journal - a proposed syslog replacement

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

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.