[coyotos-dev] Thoughts on (non)persistence
devbox at selnet.org
Tue Feb 8 13:21:54 EST 2005
On 08/02/2005, at 11.38, James Graves wrote:
>"Jonathan S. Shapiro" <shap at eros-os.org> wrote:
>> Then you aren't solving the right problem. I wasn't asking for
>> consistency here. I was asking for a means for a server that manages its
>> own consistency to recover state after it resumes. Even if it is not
>> done transparently, we still need a means for per-process *explicit*
>> state recovery.
>So hopefully the big win here is to have the OS provide good mechanisms
>for processes to recover their own state. And because they are
>standard, and always available, hopefully the application writers will
>use them in preference to creating their own state recovery /
>This will hopefully be a better situation than having all application
>writers create their own recovery mechanisms on top of the filesystem
>(or a database). Even though the filesystem is the most common
>persistence mechanism, it doesn't provide any structure, so we have a
>variety of configuration file formats, for example.
>Is this correct?
This is exactly what I meant when talking about "uniform storage model".
More information about the coyotos-dev