Help needed to convince management...

[PJ on Development]

Hi All...

I needed to convince my manager that his view of the way to implement
WebServices is a little bit off mark.

First let me explain the problem.

Currently we have quite a number of legacy web application (ASP/CGI)
that we're planning on rewrite as WebServices -- and I couldn't agree
more.

Now that's the problem. In my manager's view he thinks that each
command should be a differencet WebService.

For example, if we provide a service to, say, list and add people to
our extension directory, he wants a WebService to list and another one
to add.

I *MY* view I'd have a single WebService that provide methods to list
from and add into our extension directory.

But I can't translate the dificulty I see when people will be
consuming this WebServices into proper words.

So, I'm asking you guys, for some help to translate this into proper
managerial wording.

Regards,

Paulo Santos
http://pjondevelopment.50webs.com

 

[AMercer]

Now that's the problem. In my manager's view he thinks that each

command should be a differencet WebService.
I *MY* view I'd have a single WebService that provide methods to list
from and add into our extension directory.

If you have a small number of developers, I see your point (sort of), but if
you have many, your manager may have it right because the project breaks up
naturally into pieces for independant development requiring (relatively) less
coordination.

For every well run project, there will be a regression test policy once you
get to O&M. In your approach, the smallest delivery will require a complete
regression test (or you can take short cuts), whereas your manager's approach
will require limited regression tests.

But I can't translate the dificulty I see when people will be
consuming this WebServices into proper words.

What difficulty? Efficiency maybe? Code duplication vs reuse? I think the
burden is on you to explain your problem a little more. So far, I don't see
any compelling argument in either direction.

 

[PJ on Development]

Hi, first of all thanks for the promptly reply.

Now addressing your points:

If you have a small number of developers, I see your point (sort of), butif
you have many, your manager may have it right because the project breaks up
naturally into pieces for independant development requiring (relatively) less
coordination.

For every well run project, there will be a regression test policy once you
get to  O&M.  In your approach, the smallest delivery will require a complete
regression test (or you can take short cuts), whereas your manager's approach
will require limited regression tests.

What difficulty? Efficiency maybe? Code duplication vs reuse? I think the
burden is on you to explain your problem a little more. So far, I don't see
any compelling argument in either direction.

Yes, we are a very small development team, I have two developers under
my supervision, a team of two testers to ensure software quality and
two designers that creates the "beauty" of the softwares.

We have structured the project in a way that each WebMethod is self
contained, i.e., it is completely stateless and deterministic (ya
know, GIGO ;-), so regression tests are fine and done automaticaly in
each nightly build.

What I see is that we have several small services (the afore mentioned
Extension Directory, a common scheduller, and a few other small
services) that I see as small individual components, that I see as
unities.

I'm not sure how to express this idea better, but the way my manager
sees, when someone consumes the WebServices they need to reference
several WebServices to call the methods related with a single task
group, when I'd rather reference a single WebService that would
provide all the methods I require to accomplish the same goal.

I think that both me and my manager are talking basically the same
think in slightly different way, because both of us do have a strong
focus on mantainability of the code generated, as well a paramount of
code quality.

Just to comment, one of our check-in policy spell check each and every
word on any string resource used by our application, against a full
blown dictionary.

Regards,

Paulo Santos
http://pjondevelopment.50webs.com

 

[Andreas Johansson]

Hi Paulo,

something that you could try is to demonstrate the overhead that comes with
using many different webservices that are related to one task. If you can
show the extra costs in traffic and time needed for a client to complete
tasks it will be one argument to avoid splitting it up over multiple
services.

Some prototyping and benchmarking and you should be able to get some
measurement to show.

Best regards,
Andreas