Standardizing Batch methods?

Hello,

are there any plans for standardizing Batch methods in
WebDAV? Is there
a spec that could be discussed or are there real
alternatives to
transactions in WebDAV?
I searched in the WebDAV mailing list archive, but the last
posts
according to that issue are quite old. 

Cheers,
Markus
-- 
Markus Litz
DLR (German Aerospace Center), 
Simulation and Software Technology
Linder Hoehe, 51147 Cologne, Germany
voice: +49 2203 601 2018 	fax: +49 2203 601 3070
eMail: Markus.Litz at dlr.de      http://www.dlr.de/sc

markus.litzdlr.de wrote:
> Hello,
> 
> are there any plans for standardizing Batch methods in
WebDAV? Is there
> a spec that could be discussed or are there real
alternatives to
> transactions in WebDAV?

I'm not aware of any plans. Things I *am* aware of are:

- two documents for which documentation was requested (BIND,
SEARCH) -- 
waiting for the IESG

- a proposal to work on an equivalent of CaldDAV for vCard.
Potentially 
a WG will be formed for that, and there was some discussion
about 
starting some separate documents simplifying this in the
WebDAV stack 
(such as defining a request body for MKCOL, or extrating
stuff like 
REPORT from RFC3253).

- maintenance of existing specs (RFC3253bis, RFC3744bis,
RFC4918bis...).

> I searched in the WebDAV mailing list archive, but the
last posts
> according to that issue are quite old. 

I would guess we'd need to come up with a problem statement
first.

There are both batch calls and transaction support in
Microsoft 
exchange, and as far as I recall these have the following
issues:

Batch: underspecification, not clear whether it's good for
retrieval 
because it bypasses caches.

Transactions: breaks GET, because specifying the transaction
id as a 
request header essentially makes all requests
content-negotiated.

So... what's the problem you're looking to solve?

Best regards, Julian

On Oct 5, 2007, at 10:15 AM, <markus.litzdlr.de> wrote:
> are there any plans for standardizing Batch methods in
WebDAV? Is  
> there
> a spec that could be discussed or are there real
alternatives to
> transactions in WebDAV?

DeltaV activities may be able to do what you're looking
for.

-Tim

 
We develop a WebDAV client specialized for organizing
scientific data and one of its main requirements is absolute
data integrity. There are many situations on which one user
action results in several webdav-request. This leads to two
serious disadvantages. First, if a user action leads to 50
or 100 webdav request, depending on the network bandwidth
and server performance, this could be really slow. And
secondly if the client crashes in the middle of a difficult
job, this could result in inconsistent data.
Some time ago, there was a discussion about microsofts batch
methods and transactions, which deals about exact the same
problems we facing here. So, I'm interested if in the
meantime one of this solutions had lead to a draft status or
if this issue had been discarded. Maybe our organization
could help working to accelerate the progress of writing a
draft.

-Markus


________________________________

Von: Tim Olsen [mailto:timbrooklynpenguin.com]
Gesendet: Sa 06.10.2007 15:01
An: Litz, Markus
Cc: w3c-dist-authw3.org
Betreff: Re: Standardizing Batch methods?




On Oct 5, 2007, at 10:15 AM, <markus.litzdlr.de> wrote:
> are there any plans for standardizing Batch methods in
WebDAV? Is 
> there
> a spec that could be discussed or are there real
alternatives to
> transactions in WebDAV?

DeltaV activities may be able to do what you're looking
for.

-Tim

markus.litzdlr.de wrote:
>  
> We develop a WebDAV client specialized for organizing
scientific data and one of its main requirements is absolute
data integrity. There are many situations on which one user
action results in several webdav-request. This leads to two
serious disadvantages. First, if a user action leads to 50
or 100 webdav request, depending on the network bandwidth
and server performance, this could be really slow. And
secondly if the client crashes in the middle of a difficult
job, this could result in inconsistent data.
> Some time ago, there was a discussion about microsofts
batch methods and transactions, which deals about exact the
same problems we facing here. So, I'm interested if in the
meantime one of this solutions had lead to a draft status or
if this issue had been discarded. Maybe our organization
could help working to accelerate the progress of writing a
draft.

Markus,

the main issue here is that it's totally non-trivial to
define batch and 
transactions methods over HTTP.

- for batch: things that bypass caches and pipelining may be
slower in 
practice.

- for transactions: I'm only aware of one implementation
(Microsoft's), 
and that one breaks HTTP semantics.

So, if you want to get somewhere somebody will have to make
a proposal 
and start work implementing it inside a server, proving that
it indeed 
works and performs well.

Best regards, Julian

Hi Julian,



so this is indeed a unsolved problem. I was wondering if
someone already started with writing a proposal for either
batch calls or transactions. Was there a discussion on which
a preference became clear? 
The best may be if I start with reading the microsoft batch
& transaction definitions, and than we consider to start
working on a draft.

Ideas?
Markus



-----Ursprüngliche Nachricht-----
Von: Julian Reschke [mailto:julian.reschkegmx.de]
Gesendet: Mo 08.10.2007 10:16
An: Litz, Markus
Cc: timbrooklynpenguin.com; w3c-dist-authw3.org
Betreff: Re: AW: Standardizing Batch methods?
 
markus.litzdlr.de wrote:
>  
> We develop a WebDAV client specialized for organizing
scientific data and one of its main requirements is absolute
data integrity. There are many situations on which one user
action results in several webdav-request. This leads to two
serious disadvantages. First, if a user action leads to 50
or 100 webdav request, depending on the network bandwidth
and server performance, this could be really slow. And
secondly if the client crashes in the middle of a difficult
job, this could result in inconsistent data.
> Some time ago, there was a discussion about microsofts
batch methods and transactions, which deals about exact the
same problems we facing here. So, I'm interested if in the
meantime one of this solutions had lead to a draft status or
if this issue had been discarded. Maybe our organization
could help working to accelerate the progress of writing a
draft.

Markus,

the main issue here is that it's totally non-trivial to
define batch and 
transactions methods over HTTP.

- for batch: things that bypass caches and pipelining may be
slower in 
practice.

- for transactions: I'm only aware of one implementation
(Microsoft's), 
and that one breaks HTTP semantics.

So, if you want to get somewhere somebody will have to make
a proposal 
and start work implementing it inside a server, proving that
it indeed 
works and performs well.

Best regards, Julian

markus.litzdlr.de wrote:
> Hi Julian,
> 
> 
> 
> so this is indeed a unsolved problem. I was wondering
if someone already started with writing a proposal for
either batch calls or transactions. Was there a discussion
on which a preference became clear? 
> The best may be if I start with reading the microsoft
batch & transaction definitions, and than we consider to
start working on a draft.
> 
> Ideas?
> Markus

I would say working on a draft can be useful, but what's
even more 
important is to find people willing to implement it. Just
writing a 
draft is unlikely to make implementations happen (even
RFC4918 IMHO so 
far has a *single* implementation...).

So I'd urge you to first understand what the problems with
the Microsoft 
implementations are, and then to check whether the advanced
features 
defined in RFC3253 do not already do what you're looking for
(at least 
with respect to transactions).

Best regards, Julian

How about starting with the workspace and activity concepts
in DeltaV?

-Joe

On Mon, October 8, 2007 10:02, markus.litzdlr.de
wrote:
>
> Hi Julian,
>
>
>
> so this is indeed a unsolved problem. I was wondering
if someone already
> started with writing a proposal for either batch calls
or transactions.
> Was there a discussion on which a preference became
clear?
> The best may be if I start with reading the microsoft
batch & transaction
> definitions, and than we consider to start working on a
draft.
>
> Ideas?
> Markus
>
>
>
> -----Ursprüngliche Nachricht-----
> Von: Julian Reschke [mailto:julian.reschkegmx.de]
> Gesendet: Mo 08.10.2007 10:16
> An: Litz, Markus
> Cc: timbrooklynpenguin.com; w3c-dist-authw3.org
> Betreff: Re: AW: Standardizing Batch methods?
>
> markus.litzdlr.de wrote:
>>
>> We develop a WebDAV client specialized for
organizing scientific data
>> and one of its main requirements is absolute data
integrity. There are
>> many situations on which one user action results in
several
>> webdav-request. This leads to two serious
disadvantages. First, if a
>> user action leads to 50 or 100 webdav request,
depending on the network
>> bandwidth and server performance, this could be
really slow. And
>> secondly if the client crashes in the middle of a
difficult job, this
>> could result in inconsistent data.
>> Some time ago, there was a discussion about
microsofts batch methods and
>> transactions, which deals about exact the same
problems we facing here.
>> So, I'm interested if in the meantime one of this
solutions had lead to
>> a draft status or if this issue had been discarded.
Maybe our
>> organization could help working to accelerate the
progress of writing a
>> draft.
>
> Markus,
>
> the main issue here is that it's totally non-trivial to
define batch and
> transactions methods over HTTP.
>
> - for batch: things that bypass caches and pipelining
may be slower in
> practice.
>
> - for transactions: I'm only aware of one
implementation (Microsoft's),
> and that one breaks HTTP semantics.
>
> So, if you want to get somewhere somebody will have to
make a proposal
> and start work implementing it inside a server, proving
that it indeed
> works and performs well.
>
> Best regards, Julian
>
>
>
>
>

Julian, Joe, Werner,

thank for your hints. I will check advanced versioning,
maybe it
complies with our requirements. If it goes not far enough,
we may
consider to work on a draft either on transactions or batch
methods.
Then we plan anyway to do a reference implementation to
stimulate
others to follow that path, but I'm pessimistic that _one_
implementation could do that...

Anyway, I'll inform this list on out next steps, eventually
others
come to the same situation and need some inspiration.

So long
Markus


On 10/8/07, Julian Reschke <julian.reschkegmx.de> wrote:
>
> markus.litzdlr.de wrote:
> > Hi Julian,
> >
> >
> >
> > so this is indeed a unsolved problem. I was
wondering if someone already started with writing a proposal
for either batch calls or transactions. Was there a
discussion on which a preference became clear?
> > The best may be if I start with reading the
microsoft batch & transaction definitions, and than we
consider to start working on a draft.
> >
> > Ideas?
> > Markus
>
> I would say working on a draft can be useful, but
what's even more
> important is to find people willing to implement it.
Just writing a
> draft is unlikely to make implementations happen (even
RFC4918 IMHO so
> far has a *single* implementation...).
>
> So I'd urge you to first understand what the problems
with the Microsoft
> implementations are, and then to check whether the
advanced features
> defined in RFC3253 do not already do what you're
looking for (at least
> with respect to transactions).
>
> Best regards, Julian
>
>
>