libUCL / UCL as FreeBSD config question
Dan Partelly
dan_partelly at rdsor.ro
Sat Nov 21 08:06:53 UTC 2015
I gather then that concurrency was deemed s not important , probably not even mentioned in those talks.
I think you guys should think more at concurrency issues, and give them first class citizen status.
Please do not rush a solution.
It;s not about making sure that 2 instances of uclcomand don't overlap, it is to make sure
that **nothing** overlaps when accessing that file. No arbitrary n tools / daeomns whatever.
It is a , after all, an OS config file, not the config file of a game.
Absent the will to adopt a proper, fully transactional and atomic mechanism of storing OS configuration,
I would go back to the drawing board for a while. It may even be a long while. Please do not employ an
half breed solution to this problem/ Leave things as they are today until you figure it all out from all angles.
> On 20 Nov 2015, at 23:18, Allan Jude <allanjude at FreeBSD.org> wrote:
>
> On 2015-11-20 15:46, Dan Partelly wrote:
>> Allan,
>>
>> Thanks for clearing my confusion, and furthering my understanding on whats cooking on this front.
>>
>> The tool is dandy. I have another issue I want to ask about:
>>
>> concurrency. Is there any support in either uclib and the tools like uclcmd to ensure
>> atomic access to the ucl files ? And not on advisory level, (although if utilities would respect
>> adviasory looking … it would be better than nothing). I mean something on the lines
>> of mandatory locking.
>>
>> Was the question of concurrency discussed ?
>>
>> Dan
>>
>>
>
> Most of the discussion centered around the design of the config files,
> and the library. My tool is in the early stages and was only briefly
> discussed with the goal of showing the power of UCL from an automation
> standpoint. Obviously uclcmd can use locking to ensure that two
> instances do not overlap. Updates to the file would also be atomic (save
> to tmpfile then rename into place), and it could check that the
> modification date of the file has not changed since it was read, to
> avoid overlapping any other access to the file.
>
> In the end, I picture it being somewhat like 'vipw'
>
> --
> Allan Jude
More information about the freebsd-hackers
mailing list