At the API level that most users work with caches, I think that's a bit too
far removed. But I agree, we just need stable formats for actual release.


On 25 August 2014 14:56, Paul Benedict <[email protected]> wrote:

> I wouldn't even care about a stable format. I'd take an all or nothing
> approach to caches. Either you can correctly read out the objects, or throw
> away your storage upon error and rebuild from scratch.
>
>
> Cheers,
> Paul
>
>
> On Mon, Aug 25, 2014 at 2:45 PM, Gary Gregory <[email protected]>
> wrote:
>
>> Right now I'm interested in making the objects serializable at all in the
>> first place. I think we can worry about a stable format once it works ;-)
>> this is WIP.
>>
>> Gary
>>
>>
>> -------- Original message --------
>> From: Matt Sicker
>> Date:08/25/2014 14:59 (GMT-05:00)
>> To: Log4J Developers List
>> Subject: On serializability
>>
>> I think it's great to make things Serializable as it helps move objects
>> around. However, we should try to come up with some stable serialization
>> formats. I don't know how many times I've had to diagnose cache-related
>> problems due to serialization formats changing. Personally, I use a
>> serialization pattern based on Externalizable and I always end the object
>> stream with a boolean to indicate whether or not there is more data
>> available (i.e., extended versions). I'm not sure how well this would work
>> in a public API, but it's worked rather well in practice so far in
>> preventing incompatible class errors and the like.
>>
>> --
>> Matt Sicker <[email protected]>
>>
>
>


-- 
Matt Sicker <[email protected]>

Reply via email to