Re: [dsml] Re: File binding straw man

[rowley@netscape.com (Pete Rowley)]

Title: Message

I dont want to flog a dead horse, so Ill respond to these points and leave
it at that.

Jeff Parham wrote:

4AEE3169443CDD4796CA8A00B02191CD03535B88@win-msg-01.wingroup.windeploy.ntdev.microsoft.com">

I view the correlation of the protocol and the data format as a benefit of DSML 2 rather than a drawback -- the grammar makes it clear what the data represents rather than leaving it to interpretation.

 

I do not disagree that it is useful.  I just disagree that it is always useful.

4AEE3169443CDD4796CA8A00B02191CD03535B88@win-msg-01.wingroup.windeploy.ntdev.microsoft.com">

Defining DSML 2 transports that spout DSML 1 grammar seems wrong to me.  DSML 2 is a substitute for DSML 1, not an addition to it.

 

DSML1.0 is a portable data format.  That is, it is portable not only in the
normal sense, but also that it is protocol and operation independent - it
is a data representation.  DSML2.0 is currently a protocol specification,
data representation is expressed in terms of the protocol only.  These are
two different things, DSML1.0 is to the current DSML2.0 draft as LDIF (well,
a sub-set anyway) is to LDAP.  For DSML2.0 to replace DSML1.0, it should
also (IMHO) provide for a portable data representation.  Since I don't believe
in re-inventing the wheel when the wheel is round enough, mimicing DSML1.0
seems appropriate in some cases.

4AEE3169443CDD4796CA8A00B02191CD03535B88@win-msg-01.wingroup.windeploy.ntdev.microsoft.com">

Per my previous mail, it makes more sense to me to define features in the file binding that simplifify the translation of search results into add operations -- e.g., by mandating the capability to transform one representation to the other on input or output.

 

Providing for semantics to be expressed in a protocol and then to explicitly
provide for ignoring them seems odd, to me at least.
      
Pete

4AEE3169443CDD4796CA8A00B02191CD03535B88@win-msg-01.wingroup.windeploy.ntdev.microsoft.com">

-J

-----Original Message-----
From: Pete Rowley [mailto:rowley@netscape.com]
Sent: Wednesday, October 10, 2001 12:33 PM
To: Jeff Parham
Cc: Shon Vella; dsml@lists.oasis-open.org
Subject: [dsml] Re: File binding straw man

Jeff Parham wrote:

4AEE3169443CDD4796CA8A00B02191CD02BDFE98@win-msg-01.wingroup.windeploy.ntdev.microsoft.com" type="cite">

On the otherhand if you would like to submit a different proposal for the file binding or an alternative change to the existing proposal, please do.

So, I slept on this.  I think the problem is that currently the representation of the data is inextricably intertwined with the mechanics of the protocol.  For other bindings this may not be such an issue, but for a file binding it seems more limiting.  So, in an attempt to extract protocol mechanics from data representation and formalise that process - how about adding an optional URI which directs search result sets into another file which is to be in dsml1.0 format?  Of course, to make that useful, similar functionality would be needed for add operations.  Make sense?

Formalising the way to achieve a server to server data dump seems more useful than requiring that "servers"  (of any kind) deviate from the standard in order to provide everyday functionality.

-- 
Pete Rowley
Developer
Netscape Directory Server

-- 
Pete Rowley
Developer
Netscape Directory Server