[infinispan-dev] [ISPN-78] Feedback needed

[Manik Surtani]

On 15 Apr 2011, at 17:53, Olaf Bergner wrote:

> Am 15.04.11 17:01, schrieb Manik Surtani:
>> On 10 Apr 2011, at 21:07, Olaf Bergner wrote:
>> 
>>> Keep in mind that so far I have completely ignored the issue of
>>> supporting transactions when reading and writing large objects. I would
>>> prefer to have a working core implementation before tackling the more
>>> complicated aspects.
>> 
>> How do you maintain consistency without transactions?  E.g., Concurrent readers and a writer?
>> 
>> Concurrent writers isn't a problem since we don't support this, however we may be able to add some kind of concurrent write support if we consider the streams as append-only.
> I never meant to actually *publish* large object support without 
> mechanisms ensuring consistency in place.

Of course - I wasn't suggesting that stuff was fully baked, I was just curious as to your plans on next steps, etc.  :-)

> Yet I would prefer to have 
> most of the other issues - what should the official API's first 
> iteration look like?, is the approach I've taken so far basically sane? 
> and so forth - sorted before moving on to the more complicated aspects. 
> So far, it took some time to get used to the code base and acquire 
> *some* knowledge about INFINISPAN's inner workings, but it hasn't been 
> exactly rocket science. Just a lot of work. Ensuring consistency, 
> however, won't probably be that easy. Especially since I only have a 
> very shallow understanding of INFINISPAN's transaction support's inner 
> workings.

Mircea will be able to help you on this.  I think he has some slides on the subject that he's preparing to present at JUDCon - I'm sure he'll be happy to share them with you.  :-)

Cheers
Manik
--
Manik Surtani
manik at jboss.org
twitter.com/maniksurtani

Lead, Infinispan
http://www.infinispan.org