Gianugo Rabellino wrote:

> Mark J. Stang wrote:
> > There are several people who are using the embedded driver
> > on this list.   I wrote an XML Server.   I can send it requests
> > like "select customer from customers where lastname = 'Stang'".
> > It is a stand-alone process.   It communicates via JMS.
> > So, my Clients "publish" requests for data and my server
> > "subscribes" to those types of requests.   When a request
> > comes in, it processes it and "publishes" the response back
> > to the clients.
>
> I have never heard of such a kind of inversion of control... client
> publishing and server subscribing: quite funny, but if it works for you,
> it's OK.

Actually, they both Publish and Subscribe.   The Server subscribes to
requests for
data.  The Clients Subscribe to changes to their data.   They communicate
via
Publish/Subscribe.

>
> > I decided to have a single process non-server version
> > of my application.  By using the embedded version in
> > my client, I can remove all the other server processes.
> > This is great for demo programs and for Customers
> > who might want to try before they buy.   I can send
> > them the embedded version/single user and they can
> > key in real data and see the application in action.
> > If they decide to buy, I ship them the server jars,
> > start the server processes using the data from
> > the embedded version and they don't loose anything.
> >
> > With the single user version, I can "upgrade" my
> > customers to a multi-user version without having
> > to re-install the entire application.   I change one
> > jar on the client, install the server, move the
> > db directory and I am done.
>
> OK, until now you are making a technical point from a convenience
> functionality. Which is fair, but IMHO doesn't really pay off from what
> we are losing from an architectural and developer POV: we have to
> maintain two drivers, keeping them in sync and evolving them at the same
> pace, which requires a lot of effort. Way too much, for having just the
> commercial convenience of being able to switch drivers and version
> between a demo and an uncrippled version (true, this convenience would
> apply to other scenarios as well). But the embed driver is here and this
> is not really a choice anymore, at least for 1.x, so enough ranting.
>

I think the bottom-line is that the embedded version of Xindice is available

to build stand-alone applications or other servers.   As such, it is
valuable
to many members of the Xindice community.   I am very disturbed that
you see a future where it doesn't exist.   Like it is a bug you plan to
remove
in the future.   IF THAT IS THE CASE SPEAK UP NOW AND I
WILL FIND ANOTHER DB.   SO PLAN ON LIVING WITH IT
FOR 1.X, 2.X, ETC.   And no, my caps lock key is not stuck
in upper case.

>
> > My only concern is that Xindice gets an
> > extra layer on top of what is already there
> > that I won't have any use for.   I would like
> > to see the embedded driver be able to bypass
> > the security.   I think it should either be an
> > API on top of the embedded driver.
>
> There are no API to invent, actually. We already have the only API we
> need (well, not really, let's say it's the only one we can use) at the
> XML:DB level, which is Database.getCollection(collection, user,
> password). Other than that, I'm not that much willing to go with the
> official distribution: of course there will be some possibility to hack
> your way through the AAA code (which, if will be JAAS based, will be for
> sure flexible enough), but I don't see this as an option for the
> mainstream development.
>
> Ciao,
>
> --
> Gianugo Rabellino

Ciao,
Mark

--
Mark J Stang
System Architect
Cybershop Systems

begin:vcard 
n:Stang;Mark
x-mozilla-html:TRUE
adr:;;;;;;
version:2.1
email;internet:[EMAIL PROTECTED]
fn:Mark Stang
end:vcard

Reply via email to