Print

Print


On Wed, 5 Dec 2007, Farrukh Najmi wrote:
> Dr R. Sanderson wrote:
>> On Wed, 5 Dec 2007, Ray Denenberg wrote:
>>>> Like... who is the author of a search result?
>>> And who is the author of an ATOM feed? If the ATOM standard makes author
>> The spec says:
>>     The "atom:author" element is a Person construct that indicates
>>     the author of the entry or feed.
>> So if author is the author of the item in the resultset (see also John 
>> Harrison's post) is that the administrator? the searcher?
>> the software? the software's developer? None of which are very useful.
>
> Or it can simply be "Unspecified" or "Unknown".

There's going to be a LOT of boiler plate text just for the alleged 
advantage of using this spec.

In the end, it's going to look like:

<entry>
   <id>Unknown-[timestamp]</id>
   <author>Unknown</author>
   <title>Unknown</title>
   <updated>Unknown</updated>
   <rights>Unknown</rights>
   <content>
     <sru:record>
       <sru:recordIdentifier>123455</sru:recordIdentifier>
       <sru:recordFormat>xml</sru:recordFormat>
       <sru:recordData> [data here] </sru:recordData>
       ...

So where's the advantage again?  The consumer still has to know all of 
the sru namespaced elements, AND it has to check and throw away all of 
the ATOM elements.

Or the simpler option is to not return the ATOM wrapper.


> Why are we so concerned with the minor issue of overuse of mandatory 
> attributes in a small set of cases?

Because it's not a minor issue, and it's not a small set of cases.
[And they're not attributes, they're elements]
None of the elements in ATOM seem to match up to the elements we have in 
SRU.  And if it's /mandatory and default/ then it's EVERY case.

You could equally well say that we should use <ol> and <li> elements ... 
as they're much more commonly understood than ATOM, and at least have 
the right semantics!


> It seems to me like a red herring. We can treat these isolated cases as spec 
> bugs and simply provide guidance on how to cope with them
> when the information is not available. Lets not throw the baby out 
> with the  bath water.

I have yet to see where the baby is in this bath?

> ATOM 1.0 is an IETF standard and has mass market adoption because of its 
> simplicity and extensibility.

XHTML is a much more widely adopted standard. Hence I propose that we 
should instead treat this as a micro-format or profile of XHTML.


> These minor issues are being worked on in the next version of ATOM where we 
> could provide useful and constructive input based upon practical experience.

Fantastic, you mean we get to do all this again when ATOM changes?


> However, none of this justifies reinventing the wheel with our own format.

Since we have our own format that predates (IIRC?) ATOM, we're not 
-reinventing- anything.  Oh, and RSS vs ATOM? Come on now, that's the 
definition of reinventing the wheel.


> I am sure that we can find quite a handful of warts in the current SRU 
> response format (e.g. lack of even an option if author etc.)

There's space for arbitrary metadata in extraRecordData and 
extraResponseData.  We can't predict what metadata will be available or 
useful, so we don't try to mandate anything.


> Frankly I see missing these important attributes (particularly id) in an 
> information management context to be sacrilege. Having
> these as required may be over kill but is is certainly better then underkill.

SRU 1.2 has recordIdentifier.  Or did you mean 'important' attributes 
like <title> and <author> of non bibliographic data?


>> Or ... we could just say NO to mandatory ATOM.
> This whole thread has taken a negative and adversarial stance towards ATOM 
> starting with its title and tone. Why?

Personally, because I haven't seen any advantage in specifically using 
ATOM, and a lot of disadvantages.  The only alleged advantage is that 
it's better understood (I debate the point, given the number of 
different semantics for the mandatory <id> element so far) and that the 
tools will deal with it natively, which I also debate given that we're 
going to have a lot of template text and then a bunch of namespaced 
elements that the tools will never have seen before.


> No spec is perfect. ATOM is only going to be better, more in demand and more 
> ubiquitous over time.

Then lets wait until then before we make it the /mandatory and default/ 
response type.  I'm happy with it being an /option/ just not 
mandatory and not the default one.


> I suspect that if this issue was put in front of open-minded end-users then 
> ATOM as mandatory and default response format would win hands down
> over *ANY* other format.

I suspect that XHTML would win, actually.

How about you generate a full ATOM based response. I'll do the same for 
an XHTML equivalent. We'll give it to people with a one paragraph 
explanation and see which is preferable.
But you can't subvert the ATOM semantics or requirements and you have to 
maintain the capabilities of SRU. (And I have the same restrictions with 
XHTML, of course)

We can also put the responses into a bunch of tools and see what happens.

Okay?

Rob