Hi Julian and all others ...
I would like to add another thing:
It would be great if we could also provide something I would call "Agents".
Currently PLC4X support talking to PLCs in a protocol they already support.
Yesterday I had a chat with guys from Phoenix Contact which also have an
AppStore-like system, where customers can install modules to the PLCs. These
modules can have full access to the PLCs shared memory. I would like to try out
creating a native C++ based App for the PLCNext PLCs which communicates via
Thrift to send "native PLC4X" requests over the wire. The counterpart on the
client side would be a "plc4x" driver ... I would also like to call that
protocol "plc4x", "agent" or "plc4x-agent". All sorts of PLC4X agents would use
the same driver as client.
At least I know the Codesys PLCs also support this installation of modules,
however I think there are windows and Linux versions ... so a little more
difficult.
Here I guess we could get the maximum performance and the maximum integration
into PLC4X.
Chris
Am 17.04.19, 09:06 schrieb "Julian Feinauer" <j.feina...@pragmaticminds.de>:
Hi all,
as we had a lot of non-technical discussions and topics the last time (the
coming of age of a software project, I guess) it’s time for us to go back to
the real fun part and do technical shit.
I had a lot of discussions (on list and off list) with several people like
Chris, Matthias, Björn, Tim and others and wanted to share my thoughts on the
future of PLC4X as I see it (from a solely technical perspective).
Currently, I see several “fronts” or centers of activity (or where I think
we should spend it).
* Language adoption – We should define and deliver APIs and bindings
for other languages to bring what we currently have to other people and other
communities. The activities we have there are currently (from my head): Markus
and C++, Björn who wanted to investigate C# and the “Interop Server” which I
played around a bit (in fact, Matthias made a python binding yesterday…)
* Driver Generation – This is a well-known Topic which is currently
driven by Chris. This is a large topic, which includes
* Model Generation (currently dfdl and state-xml)
* Templates for many languages (will partially derive from above)
* A build process, to wire both together
* Some kind of Test Suite to check the correct generation of drivers
* Automated Documentation / Spec Generation (!!
* Ecosystem / Tools – We have a set of tools that are based on PLC4X
and which enable to do things which where unthinkable before. Some are
* Scraper – A tool to scrape massive amounts of data from multiple
PLCs based on a yml configuration, this is mostly driven by Tim
* OPC UA Server – Yet to come. Maps OPC UA requests to PLC4X
requests which then go native to the PLCs. Matthias started some work on this,
Tim looked over it and I think Chris plans on implementing something here also
* We had multiple discussions about tools that “guess” something
about locations of variables or their types. Chris brought that up yesterday
and plans to do something there, Matthias and I discussed this several times
and we plan to also do something with one or two students there
* New programming models – As plc4x is open, it allows us to implement
new programming models on top of it. The best example I can give is OPM, the
JPA equivalent of PLC4X. The idea is to work with POJOs and annotations and
EntityManagers (as Beans) and have a “type safe” and Business-esque way to
communicate with PLCs.
Here I see a lot of potential and possible next steps could be (discussed
by Matthias and me)
* “Richer” Typesystem (not just primitives and Arrays as currently)
which covers complex objects
* Mapping of complex objects from POJOs to PLC segments (Like
structs in S7 or ADS)
* Auto-generation of annotated POJOs from PLC programs (much like
JPA or the C# ORM does that based on an existing database). This could be a
“killer-feature” as it would really allow type-safe end to end communication
with the plc with zero plc specific knowledge
Other Topics in this area that can be named are
* A connection pool to share / reuse connections for efficiency
(which was implemented by Sebastian and is absolutely crucial for us!)
* A central monitoring component (similar to how a Webserver
monitors each side access and the results and latencies and so..), I am
currently working on this and hope to provide a PR soon
Of course, all of this is solely based on my personal opinion or things
that came out in discussions with other involved people.
For me, this structure makes sense and perhaps it helps us to “broaden” our
scope a bit from the initial focus (drivers, drivers, drivers) to the new
picture which evolved over the last to years.
Of course, feel free to agree, disagree or participate with other opinions.
Julian
PS.: I could offer to bring this in a more “presentable” form and prepare a
short “overview” talk about this for the next meetup, if interesting
