Re: More on Paging

On May 3, 2014, at 10:02 AM, ashok malhotra <ashok.malhotra@oracle.com> wrote:

> As you can imagine, several Oracle products are struggling
> with paging. I had a chat with a couple of those folks and
> they had some interesting insights based on several
> interfaces they have studied.
> 
> 1. Next/Prev URIs are not necessary.  The client can just
>    send a sendNext or sendPrev command and the server can
>    decide what to send.  Or, the client can request a range 
>    of items based on a URI template.

I understood part of the point of the Next/Prev URIs to be 
that this is how the client knows that what it's received is 
partial -- that there *is* a Next/Prev to request.

Further, without these URIs, we run once again into the concern
you've raised a number of times -- what happens when a change
during paging moves what would have been on the Next page into
what would have been on the Current page (had that change been 
made before you requested the Current page), and you request 
Next?  How does the server know to start the Next page with
that now-relocated item, for you?


> 2. On closer examination, snapshots are more expensive than
>    I had originally thought. If the client really, really
>    wants consistency we can provide commands to create a
>    snapshot and then page from that snapshot.

And we turn again toward reinventing isolation & cursoring...


> 3. A better alternative is for the server to cache the
>    collection and then page from that cache.    Clearly,
>    not an option for huge collections.

I don't see much if any conceptual difference between such
caching and the snapshots in #2.  Yes, I think neither should
not be mandatory nor universal, but they should at least be
discussed as a future-focused option, and the aim should be
toward a negotiated client/server handling.


> 4. It is quite difficult to create eTags for the collection
>    to check if it has been modified while paging thru it.    
>    We discussed various database techniques that may help
>    such as System Commit Number but none seemed to offer
>    a good solution.

This seems an argument *for* snapshot/cache of collections,
but again, I'm sure that mandating such will not fly (even 
if most/all implementations eventually include the option).

Regards,

Ted




--
A: Yes.                      http://www.guckes.net/faq/attribution.html
| Q: Are you sure?
| | A: Because it reverses the logical flow of conversation.
| | | Q: Why is top posting frowned upon?

Ted Thibodeau, Jr.           //               voice +1-781-273-0900 x32
Senior Support & Evangelism  //        mailto:tthibodeau@openlinksw.com
                             //              http://twitter.com/TallTed
OpenLink Software, Inc.      //              http://www.openlinksw.com/
         10 Burlington Mall Road, Suite 265, Burlington MA 01803
     Weblog   -- http://www.openlinksw.com/blogs/
     LinkedIn -- http://www.linkedin.com/company/openlink-software/
     Twitter  -- http://twitter.com/OpenLink
     Google+  -- http://plus.google.com/100570109519069333827/
     Facebook -- http://www.facebook.com/OpenLinkSoftware
Universal Data Access, Integration, and Management Technology Providers

Received on Monday, 5 May 2014 13:41:40 UTC