← Back to team overview

openerp-community team mailing list archive

Re: proposal to discuss forking of OpenERP GTK client

 

2012-10-12 12:25, Fabien Pinckaers skrev:
Hello,

I am afraid that dropping the official support for the GTK client means
that the RPC/webservice API will degenerate and the main development
will diverge.
Of course, we will keep maintaining the RPC/Web-service. It's very
important as OpenERP is often integrated with third party applications.
It's a very good feature of OpenERP.
As long there are no official libraries and no official "second" client, there are no way to prove that's no flaws in the API. The "good" feature will be gone long before 8.0 of OpenERP. Its not a "good" feature, its an essential feature, way more important than any jQuery "good to have" interface feature.


We are also improving the protocol to support others authentification
methods than the plain text login/password. v7.0 will support different
authentification mechanisms: oauth, openid, ldap, login/password.

There are other GTK-based application that have implemented oath, openid or even facebook authentication. Thats not a showstopper what so ever.

Even if we got a "fork" of the GTK-client there are no commitment from
OpenERP SA that it will be possible to maintain such thing.
Its too tempting to use bells and whistles from web and bypass API and
create a mess.
If we want to refocus our efforts on the web, it's not for the "bells
and whistles" as some pretend it is. For v7.0, less than 5% of our
developers worked on the new design. Most of our efforts where on making
everything robust.
Its NOT focusing on booth web client AND api-library/GTK-client at the same time that costs resources. If every step taken with the web client also where taken with api-library/GTK-client the resulting product would have been better tested, more robust and better quality. Its nearly shameful to tell the community: - We have now done a lot of work with the API, and taken a lot of API-descisions that that will not work with neither the api-library nor the GTK-client, we have spent zero time to maintain it for a long time, so if there are some one that want to spend 30 man month to do what we should have done 18 month ago you are welcome. We think you should pay the price that we had brain-fag and negledcted one of the primary functions in our product.

As a matter of fact, the cost to get on track will increase exponantial, and several steps and design descisions for the web client need to be undone, the longer this diverge will continue.


The advantages of the web client:
   - modularity
   - allows to easily develop features that can be very complex in GTK:
     touchscreen point of sale, new reconciliation facilities, advanced
     widgets like many2many_tags, kanban, good integration of external
     libs (geo-engine of camptocamp, linkedin integration, oauth, ...)
   - it will allow to go mobile.
There are no advantages with the web client if we have to rewrite it all in the next sprint.

Please don't waste that unique position OpenERP has today.
v7.0 will allow an even better unique position: ERP accessible to
everyone. You will see that the demand on OpenERP will explode and that
your implementation effort will be reduced a lot.
No there are no unique postition without working api-library/GTK-client, just another cloud service.

We dont have a proposotion for customers tired on cloud services delivering desktop applications to their employees. With a web client we have to build an UX of our own, it will look like nothing else. You cant combine it with any other software UX-like on the desktop. At the same time there are thousands of UX-experts working with desktop experience that we not will be part of. That's not optimal. I meet many young companies that have built there own infrastructure on a bunch of cloud services, that want to go from that splintered environment for there employees to a single same looking UI and integrated dataflow for all applications. With a GTK-client and api-library we can blend in and be a part of that concept.

With that said, a good web client are a nice thing but not nice enough to sacrifice api-library and GTK-client.

Regards,

Anders Wallenquist




References