Actually I think we have a problem with all three points in the CAP
theorem, so we should start coding for fault tolerance. Still this
goes off-topic. I'm done with this thread.

On Mon, Jul 13, 2015 at 6:29 PM, John Erling Blad <[email protected]> wrote:
> A property function for versioned labels would look _exactly_ the same
> as today, but the request for the actual value would use the label and
> a timestamp. The timestamp would be the last revision of the template.
>
> Look up books on distributed databases, and check out how they
> maintain consistency. In our case it is a simplified set up with a
> single master server with multiple clients that has read access to the
> server. In addition they have their own state that interfere with the
> master state.
>
> We have a consistency problem (CAP theorem) because the clients don't
> update their state (the templates) according to changes on the server
> (the labels). One solution to this is to keep the last known version
> (a timestamp) to make it possible to continue using outdated
> information.
>
> Another way to say this is that a change in the label leaves an
> invalid state at the clients, because the transaction ends prematurly.
> That is the templates are not updated, which they must be if the
> system lacks versioning.
>
> Even another way to describe this is that the process running on the
> server and the clients lacks isolation, which again can be restored
> with versioning.
>
> There are several consistency models that can be used, but I don't
> know if anyone includes something like the proposed "alias model".
>
> CAP theorem is described in these two, but I havn't read them, sorry for that!
>
> Brewer, Eric A.: Towards robust distributed systems (abstract), in:
> Proceedings of the nineteenth annual ACM symposium on Principles of
> distributed computing, PODC’00, ACM, New York, NY, USA, pp. 7–
>
> Gilbert, Seth and Lynch, Nancy: Brewer’s conjecture and the
> feasibility of consistent, available, partition-tolerant web services.
> SIGACT News (2002), vol. 33:pp. 51–59
>
> On Mon, Jul 13, 2015 at 4:21 PM, Markus Krötzsch
> <[email protected]> wrote:
>> On 13.07.2015 16:01, John Erling Blad wrote:
>>>
>>> No we should not make the aliases unique, the reason aliases are
>>> useful is because they are _not_ unique.
>>> Add versioning to labels, that is the only real solution.
>>
>>
>> Following this thread for a while, I still have no idea what this solution
>> is. Could you give an example of how the #property function in Wikitext will
>> look for this proposal?
>>
>>>
>>> There are books on the topic, and also some dr thesis. I don't think
>>> we should create anything ad hoc for this. Go for a proven solution.
>>
>>
>> Citation needed ;-)
>>
>> Markus
>>
>>
>>>
>>> On Mon, Jul 13, 2015 at 3:24 PM, Daniel Kinzler
>>> <[email protected]> wrote:
>>>>
>>>> Am 13.07.2015 um 13:00 schrieb Ricordisamoa:
>>>>>
>>>>> I agree too.
>>>>> Also note that property IDs are language-neutral, unlike english names
>>>>> of
>>>>> templates, magic words, etc.
>>>>
>>>>
>>>> As I said: if there is broad conseus to only use P-numbers to refer to
>>>> properties, fine with me (note however that Lydia disagrees, and it's her
>>>> decision). I like the idea of having the option of accessing properties
>>>> via
>>>> localized names, but if there is no demand for this possibility, and it's
>>>> a pain
>>>> to implement, I won't complain about dropping support for that.
>>>>
>>>> But *if* we allow access to properties via localized unique labels (as we
>>>> currently do), then we really *should* allow the same via unique aliases,
>>>> so
>>>> property labels can be chanegd without breaking stuff.
>>>>
>>>> --
>>>> Daniel Kinzler
>>>> Senior Software Developer
>>>>
>>>> Wikimedia Deutschland
>>>> Gesellschaft zur Förderung Freien Wissens e.V.
>>>>
>>>> _______________________________________________
>>>> Wikidata mailing list
>>>> [email protected]
>>>> https://lists.wikimedia.org/mailman/listinfo/wikidata
>>>
>>>
>>> _______________________________________________
>>> Wikidata mailing list
>>> [email protected]
>>> https://lists.wikimedia.org/mailman/listinfo/wikidata
>>>
>>
>>
>> _______________________________________________
>> Wikidata mailing list
>> [email protected]
>> https://lists.wikimedia.org/mailman/listinfo/wikidata

_______________________________________________
Wikidata mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/wikidata

Reply via email to