> On 十二月 22, 2015, 11:55 p.m., Guangya Liu wrote:
> > docs/versioning.md, line 22
> > <https://reviews.apache.org/r/41661/diff/2/?file=1174806#file1174806line22>
> >
> >     Not yours, but can you please add the following to "Terminology" 
> > section?
> >     
> >     Scheduler API
> >     Executor API
> >     Internal API
> >     Operator API
> >     
> >     I did not get any explanation for those APIs except from a slides 
> > shared by @vinod in Seattle conference 
> > http://www.slideshare.net/mesosphere/mesos-http-api
> 
> Anand Mazumdar wrote:
>     I don't think the versioning document should cater to such trivial 
> details. The Mesos Scheduler API is already linked from the Mesos homepage. 
> The Executor API should be up there soon along with others ?
>     
>     Also, this is what the description for them would look like:
>     
>     Scheduler API: API for schedulers to interact with Mesos.
>     Executor API: API for executors to interact with Mesos.
>     Operator API: API for operators to interact with Mesos.
>     Internal API: API for internal communication between Mesos Master and 
> Agent.
>     
>     ^^ Isn't this self-explantory ?

Fair enough, thanks.


> On 十二月 22, 2015, 11:55 p.m., Guangya Liu wrote:
> > docs/versioning.md, line 36
> > <https://reviews.apache.org/r/41661/diff/2/?file=1174806#file1174806line36>
> >
> >     Not clear for what does " It would be really hard to debug an error 
> > that happens with Scheduler v2 API, Executor v4 API interacting with Mesos 
> > cluster running Internal v6 API and Operator v7 API." mean? I think that 
> > even with API versioning, this problem is still difficult to debug, how can 
> > the API version help this issue?
> 
> Anand Mazumdar wrote:
>     We have a single API version across all components in Mesos. So you won't 
> ever run into the version explosion problem i.e. you have different 
> components running on different versions.
>     
>     If any of the components have a backward compatible change it would 
> result in a version upgrade for all the components. Does this help clear 
> things ?

Yes, I know what you mean, but the sentense here to me is more like to be: With 
API versioning, I can make the Scheduler v2 API, Executor v4 API interacting 
with Mesos cluster running Internal v6 API and Operator v7 API be easy to 
debug. what do you say?  ;-)

What about refine the sentense as this: 
* **Avoiding version explosion**: It would be really hard to debug an error 
that happens with Scheduler v2 API, Executor v4 API interacting with Mesos 
cluster running Internal v6 API and Operator v7 API. By putting all the APIs 
under one version, we avoid the version explosion problem.


- Guangya


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/41661/#review111689
-----------------------------------------------------------


On 十二月 22, 2015, 10:57 p.m., Anand Mazumdar wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/41661/
> -----------------------------------------------------------
> 
> (Updated 十二月 22, 2015, 10:57 p.m.)
> 
> 
> Review request for mesos, Neil Conway and Vinod Kone.
> 
> 
> Bugs: MESOS-4192
>     https://issues.apache.org/jira/browse/MESOS-4192
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> This change adds documentation on how Mesos does API versioning. It also has 
> a section:
> - On how API versioning and Release versioning are different.
> - API compatibility/upgrade guarantees.
> - Implementation Details.
> 
> Most of the information is taken from the design doc on Mesos HTTP API 
> Versioning:
> https://docs.google.com/document/d/1-iQjo6778H_fU_1Zi_Yk6szg8qj-wqYgVgnx7u3h6OU/edit#
> 
> 
> Diffs
> -----
> 
>   docs/home.md 51c19bb9d0d74698fcdda6197d32ed8f4a57d7c9 
>   docs/versioning.md PRE-CREATION 
> 
> Diff: https://reviews.apache.org/r/41661/diff/
> 
> 
> Testing
> -------
> 
> https://gist.github.com/hatred/1cc6db05d0ca51397886
> 
> 
> Thanks,
> 
> Anand Mazumdar
> 
>

Reply via email to