Hi Scott, Taher,
I think you are both right, and maybe because you are mostly working for 2
different markets or have different types of clients.
Anyway, what I mean is:
1. Form widgets are not of much use when you have to deploy a new UI for an
ecommerce or alike project (frontend).
2. They are very useful when you are working on a backend project (ie ERP part)
where people don't care much about bells and whistle (even if that's
less and less happening) but want a fast ROI ("time-to-market reasons" as
said Taher)
I don't know if Mathieu will get enough time to succeed on his project. But obviously if we had the possibility to generate RESTful web services from
OFBiz services, with the export attribute like for SOAP and RMI, then Scott's idea would be fulfilled and that would help much, not only in the UI
area of course.
Now for widgets, the form part could maybe slowly replaced by using tools like Bootstrap and Vue.js. Or the new flavor in some years and that must be
very seriously taken into account to not have to redo it again, in few years...
Jacques
Le 15/05/2018 à 12:18, Taher Alkhateeb a écrit :
Ahhh, I understand clearly now. Thank you!
So more or less, the heart of your message as I understand it is that
we should decouple the rendering of the user interface from data
fetching and manipulation. This makes perfect sense and is a good
strategy.
A bit contrary to your experience though, most of our work relies
heavily on the widget system for time-to-market reasons. It has been
immensely beneficial to get something out the door quickly. However,
of course the system falls short when it comes to heavy customizations
or the need to integrate with other systems.
So I would suggest that perhaps your comment in this thread that
"having prebuilt APIs would have reduced the workload" is applicable
in case of custom work. Otherwise, perhaps the faster route is through
the widget system. Therefore I think it is reasonable to apply both
strategies: 1) use good modern UI tools 2) build powerful flexible web
APIs. But for standard screens, I see no reason to use web service
calls instead of <action>...</action> tags to do quick and obvious
things unless perhaps you make the web API call part of the widget
system itself (also a good idea!)
Anyway, you're making me think more seriously of pushing forward the
implementation of web services, but I think introducing these
frameworks is going to be beneficial either way.
On Tue, May 15, 2018 at 12:44 PM, Scott Gray
<scott.g...@hotwaxsystems.com> wrote:
Hi Taher,
I'm simply saying that if we were to provide a complete suite web APIs to
access the full functionality of ofbiz, then the project's choice of UI
technology no longer matters so much in the grand scheme of things. No one
would be forced to live by our choice of UI frameworks because they could
build anything they liked using the APIs without ever touching the
server-side code.
Right now our data gathering logic is tightly coupled to our UI,
inaccessible to other servers and apps, the vast majority of our services
are built to be run internally by ofbiz. Without heavy modification of the
server side code, I can't build a custom SPA, I can't send orders to ofbiz
from another application, I can't really do anything without interacting
with the OFBiz UI.
The majority of the client projects I've worked on always involve a
completely custom UI and with web APIs I could pick up any flavor of the
month UI framework to build it with.
All I'm trying to add to this conversation is that it would be nice if any
UI overhaul started with making APIs available that could be used both by
our framework of choice and be externally accessible to anyone else's
framework of choice.
Regards
Scott
On Tue, 15 May 2018, 20:27 Taher Alkhateeb, <slidingfilame...@gmail.com>
wrote:
Hi Scott,
Again thank you for the input, intriguing. I'm not sure if I fully
understand though. Are you saying we can introduce web services that can
sort of do away with the widget system to code directly in html and weaving
in web service calls? How does that make coding faster? What is inefficient
in the widget system? What kind of architecture should we have in place?
What is the routing workflow that you're suggesting?
I would appreciate a bit more elaboration to get a better understanding of
your point of view since this seems to be a critical architectural
decision.
On Mon, May 14, 2018, 9:39 PM Scott Gray <scott.g...@hotwaxsystems.com>
wrote:
On Mon, 14 May 2018, 20:38 Taher Alkhateeb, <slidingfilame...@gmail.com>
wrote:
Hello Scott, thank you for your thoughts, inline ...
On Mon, May 14, 2018 at 10:45 AM, Scott Gray
<scott.g...@hotwaxsystems.com> wrote:
I think no matter what we use someone will always want something
different.
I'm beginning to lose count of the number of custom APIs I've written
over
the years to support custom UIs.
I think the bigger win would be to start providing APIs and rewriting
our
existing screens to use them. From there we could start looking at
new
UI
frameworks.
Do you mean by APIs rewriting our XSD files and model objects? Why
rewrite? Why not just enhance them for new / missing functionality?
What are the flaws you'd like to redesign?
No, I'm talking about web services. With well designed web services
custom
projects would be able to build out new user interfaces in a lot less
time
and we'd be able to poc new interfaces for the community project without
even touching the existing codebase.
Most of the projects I've worked on have needed huge amounts of UI
customization and having prebuilt APIs would have reduced the
workload
much
more than having a shinier UI that still needs to be completely
rewritten,
although I'll admit the latter would probably help the sales process.
The "shiny" part is a plus, but not the core of my suggestion. The
reasons I suggested these libraries are:
- bootstrap: the grid system, this is the cake for me. You have a
powerful responsive grid system for better layouts. The buttons,
tables and other bling bling are icing on the cake.
- Vue: The core of this technology is to allow binding of your context
model to the DOM so that you don't write oodles of JavaScript and
Jquery to create dynamic behavior. It's really old school in 2018 to
keep jumping between many pages to get something done.
Does it not worry anyone else that our service engine still only
defines
a
basic map for in/out parameters when the rest of the world is using
the
likes of swagger and restful APIs?
Of course it worries me, and if you start an initiative I will be the
first to jump in and volunteer. In fact it's on my todo list, and I
was looking at multiple options lately and I'm very attracted to
GraphQL for example because of the reduced visits to the backend.
However, I don't see this as being related to my proposal here, I'm
just setting my own priorities of what to work on next. What's wrong
with starting _both_ initiatives for that matter?
Nothing is wrong with both, but as you pointed out many discussions and
efforts have begun and then floundered. I'm simply offering some thoughts
on where I see the most potential benefit from a large scale effort.
Regards
Scott
On Sun, 13 May 2018, 06:03 Taher Alkhateeb, <
slidingfilame...@gmail.com>
wrote:
Hello Everyone,
Recently, we at Pythys had some interactions with clients, and the
user interface proved to be a sour point. It is functioning well,
but
looks too classic, too rigid, too 2000s really :) We had many
discussion and attempts in the past like [1] [2] [3] [4] [5] [6] [7]
[8] [9] [10] just to name a few all of which seemed not to follow
through.
So what is the problem? Why is this hard to get right? I'm not sure
I
have the magic answer, but it seems to me like part of the problem
is
simply .. TOO BIG
So I was thinking about a possible solution, and after some initial
research, I think maybe the solution (like everything else) needs to
be slow, incremental and evolutionary rather than revolutionary. So
I
am suggesting the following ideas to try and move forward:
- We include the assets for Bootstrap in the common theme. Bootstrap
will give us the Grid system which allows for a responsive website
that works on all devices, it will also give us beautiful widgets to
work with.
- We include Vue.js assets in the common theme. Vue.js is an
excellent
framework for creating Single Page Applications (SPAs) to give
dynamic
behavior to our pages and create ajax-heavy pages
- We slowly migrate our old CSS to bootstrap constructs. We can
begin
for example by replacing our menus, then tables, then headers, then
buttons etc ..
- We slowly introduce dynamic screens using controller logic in
Vue.js
- We slowly update our macro library to migrate to the above
mentioned
libraries and widgets
- We do all of this live in Trunk, without branching. This means
that
for some period of time, there will be transitional code (a little
bit
of bootstrap and a little bit of our current code)
We can start with an initial proof of concept skeleton, and if that
gets consensus, then we can move forward with a full implementation
in
trunk. I think this issue is many years over due. Our interface
looks
oooooooooooooold and really needs a face lift.
What do you think? Ideas? Suggestions?
[1] https://s.apache.org/rf94
[2] https://s.apache.org/g5zr
[3] https://s.apache.org/XpBO
[4] https://s.apache.org/YIL1
[5] https://s.apache.org/836D
[6] https://s.apache.org/DhyB
[7] https://s.apache.org/Lv9E
[8] https://s.apache.org/zKIo
[9] https://s.apache.org/D6jx
[10] https://issues.apache.org/jira/browse/OFBIZ-5840
