[occi-wg] Call to discuss OCCI attributes in HTTP headers Wed Sept 30 06:00 PDT, 9:00 EDT, 14:00 BST, 15:00 CEST

[Sam Johnston]

On Wed, Sep 30, 2009 at 3:53 AM, Gary Mazz <garymazzaferro at gmail.com> wrote:

>
> Good luck with your move today.. I hope its somewhere nice. :-)
>

That's yet to be seen - I'm unconvinced but I'll still be spending a fair
bit of time in Paris and it's got to be better than "le trou du cul de la
France" where we are now.


> I do not want to spend the time discussing a shift in direction that will
> result in making the release of this document unachievable for OGF27.
>

It's clear that you either don't understand or don't like the concept (or
both) but did you realise that the HTTP headers allow us to avoid writing
the various renderings? These are both significant work requiring serious
attention to detail (basically stuff I think we should leave to the IETF
which is why I've been churning out Internet-Drafts for these areas) and
significant risk to interoperability... ironically those who [dis]like each
of the formats will end up having to implement all of them. The text
rendering <http://occi.googlecode.com/hg/docs/occi-text.html> basically puts
HTTP headers in the entity-body anyway so it's not a huge leap of faith to
put it in the headers (where metadata arguably belongs).

Furthermore, the metadata may conflict with the resource itself (for example
a container having more/less cores/memory than the workload specifies) which
breaks our model. By jumping into the entity-body we're essentially
contracting to provide a representation and even if lossy/incomplete it must
be correct. Consider the problem of starting a read-only VM template but
being confined to Amazon-style instance types, or instances that
resize/burst dynamically based on load. This completely breaks our model and
requires us to revisit the idea of separating container from workload (as I
originally suggested) - that *is* something I'd rather avoid doing at this
late stage.

Ultimately HTTP doesn't touch the entity-body so neither should we - that
makes for a very clear demarcation point as to what is OCCI and what is left
to other standards like OVF. While this may not seem so important right now,
it is when you start looking at cloud platform constructs like queues and
structured storage (databases).

I'll do what I can to be on both 3pm and 4pm calls (meme si elle va casse
mes couilles :P).

Sam


> Sam Johnston wrote:
>
> Gary,
>
>  I'm moving to another country today so it's not particularly convenient
> for me (which is why I hadn't responded to your poll yet). I'd suggest
> adjourning such a discussion until I can explain the pros and cons of the
> three main options which will either be on the call tomorrow (assuming I'm
> not carting boxes around at the time) or some time after Thursday. I'd
> explain by email now but it's 02h30 and I've got a big day ahead of me...
> just quickly the options (bearing in mind the payload/entity-body is
> effectively opaque) are:
>
>   - use a wrapper format for the metadata like SOAP or Atom (which neuters
> many nice HTTP features like caching while adding a significant encoding
> overhead, but it does give us collections)
>  - use a separate resource for the metadata e.g. have
> http://example.com/myvm.ovf as well as http://example.com/myvm.atom,
> http://example.com/myvm.json, http://example.com/myvm.txt etc. (which is
> what we had planned until recently, but which requires us to specify
> renderings in various formats)
>  - use the HTTP headers (which avoids the whole format discussion by using
> a single, standard format which can be sent with [GET] or without [HEAD] the
> payload/entity-body)
>
>  I've spent the last week running around Europe talking about Open Cloud
> and OCCI too (CloudCamps in Edinburgh, London & Frankfurt among other
> things) so there's not many changes - at least from me - this week.
>
>  Sam
>
> On Tue, Sep 29, 2009 at 1:15 AM, Gary Mazz <garymazzaferro at gmail.com>wrote:
>
>>
>> *Although, signup for this meeting was a bit lackluster,  We have the
>> opportunity to meet  on Wed. Sept 30. 06:00 PDT, 9:00 EDT, 14:00 BST,
>> 15:00 CEST
>>
>> *cheers,
>> gary*
>> *
>>
>> Thijs Metsch wrote:
>> > Sounds interesting...but Sept 25 doesn't work for me...
>> >
>> > Gary can you Maybe create a doodle for this event?
>> >
>> > Thanks,
>> >
>> > -Thijs
>> >
>> > On Wed, 2009-09-23 at 09:01 -0600, Gary Mazz wrote:
>> >
>> >> I' like to schedule a meeting to discuss Sam Johnston's proposal to
>> move
>> >> OCCI attributes elusively  to HTTP headers.
>> >>
>> >> I'd like to schedule this conference on friday Sept 25, 2009, the as
>> >> time as the weekly call.
>> >>
>> >> We need to fully understand the mapping of the attributes to the http
>> >> protocol structures and the impacts to the specification.
>> >>
>> >> cheers,
>> >>
>> >> gary mazz
>> >>
>> >>
>> >> _______________________________________________
>> >> occi-wg mailing list
>> >> occi-wg at ogf.org
>> >> http://www.ogf.org/mailman/listinfo/occi-wg
>> >>
>>
>> _______________________________________________
>> occi-wg mailing list
>> occi-wg at ogf.org
>> http://www.ogf.org/mailman/listinfo/occi-wg
>>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.ogf.org/pipermail/occi-wg/attachments/20090930/fbf13f04/attachment.html