2011/8/2 Eric Scouten :
> It falls apart badly in a highly distributed environment where ...
>
> ...
May be a RDF storage is more reasonable for this. Operations with
atomic facts can be highly distributed. And SPARQL is similar to SQL.
--
Best regards, Alexey Pechnikov.
On Sun, Jul 31, 2011 at 12:58, Simon Slavin wrote:
These two go together. Multi-master replication (one example of which is a
> document store) is relatively easy. Datestamp every value (document) and
> whichever one has the lastest date is the one you want.
>
This is
2011/8/1 Simon Slavin :
> I'm sorry Alexey, I was trying to be funny and failed. Your question is very
> important for this situation.
Oh, I'm sorry! My english is bad by night :)
--
Best regards, Alexey Pechnikov.
http://pechnikov.tel/
On 1 Aug 2011, at 7:20pm, Alexey Pechnikov wrote:
> 2011/8/1 Simon Slavin :
>>
>> On 1 Aug 2011, at 6:56pm, Alexey Pechnikov wrote:
>>
>>> 2011/8/1 Black, Michael (IS) :
This is a side-question to this thread...but has anybody every done
2011/8/1 Simon Slavin :
>
> On 1 Aug 2011, at 6:56pm, Alexey Pechnikov wrote:
>
>> 2011/8/1 Black, Michael (IS) :
>>> This is a side-question to this thread...but has anybody every done
>>> row-level locking for edit?
>>
>> What problem are you
On 1 Aug 2011, at 6:56pm, Alexey Pechnikov wrote:
> 2011/8/1 Black, Michael (IS) :
>> This is a side-question to this thread...but has anybody every done
>> row-level locking for edit?
>
> What problem are you solving?
Please stop asking key questions.
Simon.
2011/8/1 Black, Michael (IS) :
> This is a side-question to this thread...but has anybody every done row-level
> locking for edit?
What problem are you solving?
--
Best regards, Alexey Pechnikov.
http://pechnikov.tel/
___
m: sqlite-users-boun...@sqlite.org [sqlite-users-boun...@sqlite.org] on
behalf of Roger Binns [rog...@rogerbinns.com]
Sent: Sunday, July 31, 2011 10:13 AM
To: sqlite-users@sqlite.org
Subject: EXT :Re: [sqlite] Will SQLite supports UnQL?
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Any database syste
On Jul 31, 2011, at 12:58 PM, Simon Slavin wrote:
> These two go together. Multi-master replication (one example of which is a
> document store) is relatively easy. Datestamp every value (document) and
> whichever one has the lastest date is the one you want.
I hear that a lot, but
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/31/2011 12:58 PM, Simon Slavin wrote:
> Datestamp every value (document) and whichever one has the lastest date is
> the one you want.
That isn't always the correct approach and requires clocks to be
synchronized - something that is
On 31 Jul 2011, at 8:50pm, Roger Binns wrote:
> Multi-master replication is always hard
> which is why it is so rare.
>
> The simplicity of the document model is because you put everything relevant
> into a single item which typically matches the programmer's idea of an object.
These two go
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/31/2011 11:53 AM, Alexey Pechnikov wrote:
> A normalized database can't be easy replicated and checked.
That is what makes these databases (variously document oriented or JSON) so
attractive. There is no rigid schema and no two documents need
Roger, thanks for good explanation and examples of the multi-master
replication problems.
But I think a conflict-resolutions strategies may be independent. And
checking logical data
integrity is needed for systems without replication too. This checking
is not a part of the
replication but it's
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/31/2011 07:49 AM, Simon Slavin wrote:
> neither of them can enter the new price into the system, they have to each
> enter the price change,
You cannot get conflicts if both "users" are talking to the same server - I
can explain why if you
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Any database system can be nonsensical. If you have multiple people editing
prices for the same item even with a single instance database you can have
the wrong final answer. But that is a procedural and business issue, not a
technological one.
On 31 Jul 2011, at 3:42pm, Roger Binns wrote:
> Scroll down to the bottom third of this document to the Conflicts section
> where you get blurb (but not technical details) about it:
>
> http://couchdb.apache.org/docs/overview.html
Okay. So CouchDB requires intentional data entry and uses
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/30/2011 11:33 PM, Alexey Pechnikov wrote:
> You write extra code to find and resolve CouchDB replication conflicts but you
> are speak about "do not have to write extra code". One of these
> approvals is false!
Not quite and it was deliberate
On 31 Jul 2011, at 8:07am, Alexey Pechnikov wrote:
> 2011/7/31 Simon Slavin :
>> Right. So if a customer orders one of my products they'll be presented with
>> two different costs and pick the one they like. Great system.
>
> You customers have edit privilegies to you
2011/7/31 Simon Slavin :
> Right. So if a customer orders one of my products they'll be presented with
> two different costs and pick the one they like. Great system.
You customers have edit privilegies to you prices?! It's not the
replication problem of cource.
As
2011/7/31 Roger Binns :
> My original point was that CouchDB has multi-master replication builtin -
> you do not have to write extra code or do anything special.
and
> You can
> programmatically find conflicts and then it is up to you how to resolve them
> since only your
On 30 Jul 2011, at 11:31pm, Alexey Pechnikov wrote:
> There are no conflicts! Two versions are two _different_ versions of the row.
> Simple copy _all_ un-exists rows by SQL query with attached database
> or simple script
> http://sqlite.mobigroup.ru/wiki?name=sqlite3-rdiff
> It can be included
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/30/2011 03:31 PM, Alexey Pechnikov wrote:
> There are no conflicts!
If you have multiple masters then incompatible changes can be made to the
same data.
> Two versions are two _different_ versions of the row.
ie you have a conflict.
> On
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/30/2011 03:12 PM, Simon Slavin wrote:
> Conflict resolution is so hard it's sometimes impossible.
CouchDB does it well. The vast majority of the time there are no conflicts.
When there is a conflict it picks one contender (something like
2011/7/31 Roger Binns :
> It takes more than a few lines of SQL. You've also got to be careful how
> you write data (result is a DAG), you need replication code and you need
> conflict resolution code.
There are no conflicts! Two versions are two _different_ versions of
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/30/2011 02:29 PM, Mr. Puneet Kishor wrote:
> could you please tell what this other "JSON database server product" was?
MongoDB.
A major difference between MongoDB and CouchDB is how they store data.
CouchDB uses an append only structure - data
On 30 Jul 2011, at 10:57pm, Roger Binns wrote:
> On 07/30/2011 01:43 PM, Alexey Pechnikov wrote:
>> Multi-master replication can be easy with correct data structures.
>
> It takes more than a few lines of SQL. You've also got to be careful how
> you write data (result is a DAG), you need
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/30/2011 01:43 PM, Alexey Pechnikov wrote:
> Multi-master replication can be easy with correct data structures.
It takes more than a few lines of SQL. You've also got to be careful how
you write data (result is a DAG), you need replication code
On Jul 31, 2011, at 6:04 AM, Roger Binns wrote:
> A different JSON database server product used 4GB and 6
> minutes and I never bothered optimising for it. Perhaps more shocking is
> that the other product would usually answer unindexed queries faster than
> CouchDB did with indices, mostly
Multi-master replication can be easy with correct data structures.
As example:
CREATE TABLE test
(
id INTEGER PRIMARY KEY ON CONFLICT REPLACE,
cid INTEGER UNIQUE -- equal to test_record.id
);
CREATE TABLE test_record
(
id INTEGER NOT NULL, -- record version id
rid INTEGER PRIMARY
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/30/2011 12:24 PM, Max Vlasov wrote:
> so CouchDB will be a server solution
My (year old) experience with CouchDB has been that it is a poor database,
especially in what they consider "large" amounts of data. I was working
with a dataset of 10
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 07/30/2011 10:45 AM, Alexey Pechnikov wrote:
> Very interesting annonce:
> http://www.couchbase.com/press-releases/unql-query-language
It is rather difficult to work out from the spec pages how it differs from
regular SQL. Personally I think it
On Sat, Jul 30, 2011 at 9:45 PM, Alexey Pechnikov wrote:
> Very interesting annonce:
> http://www.couchbase.com/press-releases/unql-query-language
>
>
Thanks for the info, very interesting indeed.
After some reading at http://www.unqlspec.org I'd rather think Richard
Very interesting annonce:
http://www.couchbase.com/press-releases/unql-query-language
--
Best regards, Alexey Pechnikov.
http://pechnikov.tel/
___
sqlite-users mailing list
sqlite-users@sqlite.org
33 matches
Mail list logo