XQuery and ACID transactions WAS Re: from Where 2.0

[Daniela Florescu]

WOW ! What a long and serious response, Michael. I was expecting a  
laugh :-)

What can I say: I am a little perplex.

If such attitude is general, I will not be surprised that actually  
Cassandra or another semi-structured
  database in the same league will win and this community will die,  
still discussing 10 years from now
  if eventual consistency  and distributed architectures in the cloud  
are good ideas.

Or maybe that's exactly your point !?

Perplexed, best regards
Dana




On Mar 31, 2010, at 5:05 PM, Michael Rys wrote:

> Hi Dana
>
> I don't have a problem with setting standards in an area that has a  
> wide range of applicability and is stable.
>
> I however have the feeling that we often tend to think we are in one  
> of the top curves while in fact we are more in one of the situations  
> at the bottom of his posting. There is still lots of research to be  
> done around transaction models especially around the different,  
> newly emerging cloud implementation environments that I feel  
> standardizing it is premature. Even in the context of SQL, where  
> there exists some form of standardization of concurrency (above the  
> language level), advancements in research sometime end up delivering  
> better results than the options provided in the standard.
>
> The danger of early calcification due to premature standardization  
> together with the special situations in different implementation  
> environments (e.g., an XQuery implementation based on a cloud  
> storage server may want to only provide eventual consistency or  
> repeatable reads, but no full serializability, while an XQuery  
> implementation in an information integration environment may want to  
> do a SAGA or ML-TA based consistency model, while a native XQuery/ 
> XML database may want to select 2 to 3 consistency levels, while an  
> XQuery engine embedded inside a relational engine wants to integrate  
> into the existing transaction environment provided by the host  
> database, what is the conflict granularity level? Do we expect only  
> locking protocols? Or are optimistic protocols standardized as well?  
> Etc. pp.) indicates to me that it is not something we need to  
> standardize now. Especially since the relevant standards on which  
> this builds have not even been released yet!
>
> Best regards
> Michael
>
> PS: I think most of XQuery 1.0 we had a nice balance between the T  
> and P curve, even though we did some technical innovations around  
> mixing known and unknown types. XQuery FT we did both T and P in  
> some sense at the same time, but there has been enough T in the  
> technical area, that the innovation was more around how to integrate  
> it into the language and how to formalize its semantics. With XQuery  
> updates, we again mainly had a nice balance, especially given the  
> work done by Dare, Jonathan, Lehti, yourself, Don, Jerome, you and  
> myself and countless other researchers. I however think that we have  
> moved T and P way too close to each other with XQuery Scripting  
> extensions and starting to talk about standardizing processing  
> models and host environment aspects such as concurrency control  
> seems to be making the same mistake.
>
> PPS: This expresses my personal opinion and is in no way reflecting  
> any official statement of my employer or the XQuery WG.
>
>
>
> -----Original Message-----
> From: Daniela Florescu [mailto:http://x-query.com/mailman/listinfo/talk]
> Sent: Wednesday, March 31, 2010 4:27 PM
> To: Michael Rys
> Cc: Jonathan Robie; http://x-query.com/mailman/listinfo/talk
> Subject: Re: XQuery and ACID transactions WAS Re:  from  
> Where 2.0
>
> Michael,
>
> you seem to have a problem with the standards :-))
>
> So where do you think XQuery fits on the curve bellow?
>
> http://blogs.sun.com/jag/resource/StandardsPhases.html
>
> Cheers
> Dana
>
>
>
> On Mar 31, 2010, at 11:49 AM, Michael Rys wrote:
>
>> I would definitively leave transactional consistency levels to the
>> implementation environments and for now keep the standardization
>> bodies away from it.
>>
>> Best regards
>> Michael
>>
>> -----Original Message-----
>> From: http://x-query.com/mailman/listinfo/talk [mailto:http://x-query.com/mailman/listinfo/talk] On
>> Behalf Of Jonathan Robie
>> Sent: Wednesday, March 31, 2010 10:00 AM
>> To: http://x-query.com/mailman/listinfo/talk
>> Subject: Re: XQuery and ACID transactions WAS Re:  from
>> Where 2.0
>>
>> On 03/31/2010 12:48 PM, Daniela Florescu wrote:
>>>> But in my experience there are two classes of applications where
>>>> MarkLogic has been a necessary alternative:
>>>>
>>>> 1) .......
>>>> 2) if your needs for querying geospatial exceed the **scalability**
>>>> characteristics of other technologies
>>>
>>> That's why people yesterday told me that they use (mostly)  
>>> Cassandra.
>>> Because of scalability. That's
>>> a reason why NOT use any of existing XML databases, actually. In  
>>> good
>>> database tradition, they all support ACID transactions (as far as I
>>> know).
>>>
>>> Unfortunately, there is NO way that a database with ACID
>>> transactions, as good as it can be (even  Oracle RAC, which is
>>> probably the best) can compete in scalability with a database with
>>> "relaxed"
>>> transactional model and eventual consistency.
>>>
>>> We got to work on a transactional version of XQuery that would allow
>>> not only ACID, but also eventual consistency.
>>
>> OK, I'm listening.
>>
>> What would you change in the current Updates spec?
>>
>> Jonathan
>> _______________________________________________
>> http://x-query.com/mailman/listinfo/talk
>> http://x-query.com/mailman/listinfo/talk
>>
>>
>> _______________________________________________
>> http://x-query.com/mailman/listinfo/talk
>> http://x-query.com/mailman/listinfo/talk
>
>
>
> _______________________________________________
> http://x-query.com/mailman/listinfo/talk
> http://x-query.com/mailman/listinfo/talk