mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Joseph Wu" <jos...@mesosphere.io>
Subject Re: Review Request 37655: MESOS-3299: Add protobuf to represent time with integer precision.
Date Thu, 27 Aug 2015 16:47:30 GMT


> On Aug. 26, 2015, 10:34 p.m., Neil Conway wrote:
> >

Thanks for your comments!


> On Aug. 26, 2015, 10:34 p.m., Neil Conway wrote:
> > include/mesos/mesos.proto, line 101
> > <https://reviews.apache.org/r/37655/diff/2/?file=1049806#file1049806line101>
> >
> >     It might be useful to specify what the epoch is (e.g., seconds-since-Unix-epoch).

The `seconds` field actually just expresses "seconds" (like a `Duration` object, rather than
a `Time` object).  In the next review in this chain, we use this protobuf twice, once as "epoch
seconds" and once as a duration.


> On Aug. 26, 2015, 10:34 p.m., Neil Conway wrote:
> > include/mesos/mesos.proto, line 102
> > <https://reviews.apache.org/r/37655/diff/2/?file=1049806#file1049806line102>
> >
> >     BTW, have you considered using a simpler representation, such as just a single
int64 holding # of nanoseconds (or microseconds) since the Unix epoch? Using nanoseconds you'd
be able to express ~292 years -- maybe that's not a wide enough range?

We were aiming at something that could be used like the `struct timespec`.  So rather than
the range of times that can be represented, we were looking for usability (would you rather
think in terms of seconds or millions of nanoseconds?); with the option for extra precision.

(It would be somewhat simpler, code-wise, to just use nanoseconds.)


- Joseph


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/37655/#review96648
-----------------------------------------------------------


On Aug. 26, 2015, 3:56 p.m., Joseph Wu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/37655/
> -----------------------------------------------------------
> 
> (Updated Aug. 26, 2015, 3:56 p.m.)
> 
> 
> Review request for mesos, Benjamin Hindman, Bernd Mathiske, Ben Mahler, Artem Harutyunyan,
Joris Van Remoortere, and Vinod Kone.
> 
> 
> Bugs: MESOS-3299
>     https://issues.apache.org/jira/browse/MESOS-3299
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Instead of using doubles, seconds and nanoseconds can be represented like
> `struct timespec`, with one field for seconds and one for nanoseconds.
> 
> This will be important if frameworks need to compare times to make decisions (such as
for maintenance primitives).
> 
> Note about the naming:
> 
> * Time will conflict with the Time class.
> * Most denominations of time (Seconds, Minutes, Hours, etc) will conflict with Duration.
> 
> 
> Diffs
> -----
> 
>   include/mesos/mesos.proto 33e1b28f1ccbe227657a14395f81df20e0a9e193 
>   include/mesos/v1/mesos.proto 382b978dca769757171c5558b7f259870592c321 
> 
> Diff: https://reviews.apache.org/r/37655/diff/
> 
> 
> Testing
> -------
> 
> `make check`
> 
> 
> Thanks,
> 
> Joseph Wu
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message