plc4x-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christofer Dutz <christofer.d...@c-ware.de>
Subject Re: [DISCUSS] Add Wrappers to PLC4X Project
Date Fri, 11 Sep 2020 07:31:43 GMT
Hmm ... I usually always hit reply ... but I have seen some times the email just goes to the
sender ... strange ...

This time it didn't ...

Am 11.09.20, 09:30 schrieb "Julian Feinauer" <j.feinauer@pragmaticminds.de>:

    Forwarding it tot he list (i think chris simply hit reply...) : )

    Am 11.09.20, 09:27 schrieb "Christofer Dutz" <christofer.dutz@c-ware.de>:

        And my only concern about native-drivers is that I want to prevent us from becoming
lazy.

        I know we had the EIP and Modbus drivers by using external libs and that wasn't very
pleasant (At least not for me)
        We were relying on libs we had no control over. It got us to the point where we could
support these protocols,
        But looking back it cost us (mainly me) way more time in supporting them than it would
have cost me if we had implemented the parts we needed.

        So I would generally say:
        - If it’s a technical problem preventing us from implementing it fully in plc4x
(CAN, Profibus, ...) then this is the way to go ...
        - If it's a protocol problem (want to avoid implementing the protocol in order to
save time) ... then we should put it in the sandbox

        Chris



        Am 11.09.20, 09:01 schrieb "Julian Feinauer" <j.feinauer@pragmaticminds.de>:

            Hi folks,

            thanks for all the replies and the controversy in here shows that ist good to
discuss the matter, indeed : )

            I like Cesars way of putting it (which is pretty close to mine) that PLC4X is
a unified API.
            This is the Killer thing here.

            And what we currently do and mostly did was "PLC4X native". Because we get all
these nice benefits of controlling everyhting and doing code for all languages.

            But, we may end up in situations where there i seither no possiblity to do a native
implementation (profinet, profibus, CAN, ...) where we need some kind of special hardware
where I would like to still have it in the project.
            This would mean that there is some special setup instruction how to setup the
"native" part and then there is some glue code (could easily again be plattform independent)
like a PLC4X generic driver (@Christofer Dutz a bit like what we reasoned about with PLCnext)
which "binds" them together.

            As we are really small on man-power and maintainers I think it is a good and reasonable
way to go with a "integrate whats already there" instead of a "try to find the single silver
bullet and die in beatuy".

            So my reasoning at the moment is to place something like this wrapper as native
code in a generic section somewhere in the "PLC4X native" part, that it can be integrated.
But then host install instructions fort he "agent" or proprietary part somewhere else in an
"integration" part (not like integration into other downstream systems but integration of
other communication layers).

            Is this something which sounds acceptable for the community are there feelings
against it?

            Julian

            Am 11.09.20, 02:20 schrieb "Ben Hutcheson" <ben.hutche@gmail.com>:

                Hi,

                I agree with Chris, having new drivers in a contrib section would be a good
                idea to make it clear that it hasn't been developed as much as other
                protocols or that there is some constraint excluding it from the main
                driver section. The worst that could happen is that it gets culled
                eventually because it isn't maintained and no one else shows interest.

                What protocol is it?

                I'm assuming you've already set your mind on developing it, but something
                to consider. Is anybody else likely to want to use it to warrant spending
                time on it? Is there a spec that is published? Does the manufacturer change
                the protocol often? Can they provide information on the protocol?

                Kind Regards

                Ben

                On Thu, Sep 10, 2020 at 5:03 PM Łukasz Dywicki <luke@code-house.org>
wrote:

                > It is pretty good question to answer. Given that Julien proposal
                > includes native binary I can also add that socketcan stuff I am working
                > on .. is bound only to Linux due to native dependency in JavaCAN transport.
                >
                > From my perspective I can tell that it really depends on budget you
                > have. Our CAN journey started from CAN over Ethernet (UDP) and ended
up
                > in dark corner which some of you could observe over past weeks. We had
                > to do it, because that was customer requirement (skip vendor gateway).
                > If you don't have such requirement then I would say there is quite weak
                > case to invest in it.
                > My point comes from observations. I wrote couple of MSpec files and it
                > was indeed fun. Both profinet and lldp took me a little bit, but
                > structures I got structures and I was able to parse real traffic. What
I
                > skipped in my earlier attempts was implementation of driver api where
                > you actually start implementing necessary conversation logic. Having
a
                > fresh look from CANopen perspective - it is quite big effort which
                > requires time.
                >
                > My point outlining above is simple - as long as we have just one
                > dominant language as we have and we fail to attract more diverse pool
of
                > people to start writing conversation logic for other languages does not
                > give big benefit. I agree it is important, but it is much more necessary
                > now to support wider pool of end devices. Once we will build big enough
                > pool of people who have commercial interest in moving stuff between
                > operating systems and platforms we will be able to start joint
                > investment together. There is no better explanation than savings on
                > hardware when it comes to large scale deployments.
                >
                > From my point of view - I would welcome your stuff anywhere, even in
                > sandbox. Just remember to unify DLL loading logic. ;-)
                >
                > Best regards,
                > Łukasz
                >
                > On 10.09.2020 00:32, Cesar Garcia wrote:
                > > Hello,
                > >
                > > I think the concept of the project is clear:
                > >
                > > "PLC4X is a set of libraries for communicating with industrial
                > programmable
                > > logic controllers (PLCs) using a variety of protocols but with a
shared
                > > API."
                > >
                > > If your client allows you to publish the project in PLC4X, it is
a very
                > > important opportunity for this project to increase and share knowledge.
                > >
                > > As for DCOM, it is a reality that will be with us no less than 20
years
                > in
                > > the future due to its installed base [1]. We need to live with the
                > Windows
                > > and Linux environment for years to come, and that is a reality.
                > >
                > > As for solutions with DCOM we have [2], in my case which allows
using
                > > OPC-DA from Java, as in [3].
                > >
                > > My grain of sand
                > >
                > > 1.
                > >
                > https://opcfoundation.org/wp-content/uploads/2018/02/ARC-Report-OPC-Installed-Base-Insights.pdf
                > > 2. http://j-interop.org/
                > > 3. https://www.eclipse.org/eclipsescada/
                > >
                > > El mié., 9 sept. 2020 a las 8:31, Otto Fowler (<ottobackwards@gmail.com
                > >)
                > > escribió:
                > >
                > >>  I think this should be hosted and more importantly _maintained_
outside
                > >> the project.  If you want to add reference to it to the project
site or
                > >> something, that would be something to talk about.
                > >>
                > >>
                > >> On September 9, 2020 at 08:28:12, Stefano Bossi (
                > stefano.bossi@gmail.com)
                > >> wrote:
                > >>
                > >> Hi,
                > >>
                > >> personally I think this kind of approach will limit the usability
of the
                > >> library in a very narrow subset of uses do to the windows operating
                > system
                > >> dependency.
                > >>
                > >> I think you guys put a lot of effort in portability and small
footprint
                > of
                > >> the library and this is a great think in the industrial world
where
                > >> typically there are small PC or embedded one.
                > >>
                > >> Using the library in a small PC like a Rasperry Pi with a linux
distro
                > and
                > >> a lot of attention for the security and hardening of the environment
I
                > >> think is a pro for any industrial project
                > >> (e.g. Selinux, Firewall, minimal service installation, OSCAP
security
                > >> profile compliance, etc ).
                > >>
                > >> Evaluating the effort required in reversing the DCOM protocol
is
                > something
                > >> to be taken into consideration before integrating a windows
library in
                > the
                > >> plc4x code.
                > >>
                > >> Maybe this could be a transient solution or a way to validate
a full
                > open
                > >> source solution.
                > >>
                > >> Regards,
                > >> Stefano
                > >>
                > >>
                > >>
                > >> On 09/09/2020 13:35, Christofer Dutz wrote:
                > >>
                > >> Hi Julian,
                > >>
                > >>
                > >>
                > >> the issue I see here is that it will make the build more complex
(the
                > >> part using the wrapper will only be runnable on windows and
not sure
                > >> if the license of the wrapped DLLs would allow including them).
                > >>
                > >>
                > >> It will also eliminate the ability to auto-port the driver to
other
                > >> languages.
                > >>
                > >>
                > >> And, beyond that, it would limit these drivers to only work
on a
                > >> subset of platforms (Aka ... a Java Driver that only works on
Windows
                > >> Systems with installed subsystem for the PLC)
                > >>
                > >>
                > >>
                > >> I wouldn't want to make such a solution a first class citizen
(aka
                > >> live in plc4j/drivers) ... we could sort of start providing
some sort
                > >> of "plc4j/contrib" modules, if we have to go this path.
                > >>
                > >>
                > >>
                > >> But personally I would opt for at least having a look at the
path I
                > >> described in slack:
                > >>
                > >>
                > >>
                > >> - Use the native libs and build an application that does the
basic
                > >> interaction with the Windows DLLs
                > >>
                > >> - Use WireShark to record the communication
                > >>
                > >> - Have a look if it's not just a very small subset of DCOM that
is used
                > >>
                > >>
                > >>
                > >> Perhaps it would sort of be like using some mspec types with
a lot of
                > >> const fields to allow communication without any intermediate
DLL ....
                > >> this would make it runnable on all target platforms and auto-portable
                > >> to all target languages of PLC4X.
                > >>
                > >>
                > >>
                > >>
                > >>
                > >> Chris
                > >>
                > >>
                > >>
                > >> Am 09.09.20, 09:50 schrieb "Julian Feinauer"
                > >> <j.feinauer@pragmaticminds.de> <j.feinauer@pragmaticminds.de>:
                > >>
                > >>
                > >>
                > >>     Hi all,
                > >>
                > >>
                > >>
                > >>     wanted to bring it tot he list as we already had a discussion
in
                > >> the slack channel.
                > >>
                > >>     We have a project where we consider integrating machinery
in our
                > >> system.
                > >>
                > >>     The machinery offers an SDK for communication which is windows
                > >> only and based on DCOM.
                > >>
                > >>     Thus, the integration would be a wrapper around the SDK
with
                > >> „only“ a PLC4X „frontend“.
                > >>
                > >>
                > >>
                > >>     Personally, I consider this viable as our aim ist o have
one
                > >> interface for „everything“.
                > >>
                > >>     Nonetheless, I also agree with everybody saying that its
not as
                > >> nice as having the complete stack in our hands.
                > >>
                > >>
                > >>
                > >>     What do others think, should this be part oft he PLC4X project
or
                > >> should we just do it separately.
                > >>
                > >>     Personally idk that much but think it would be nice to have
                > >> maximum support in plc4x, if possible.
                > >>
                > >>
                > >>
                > >>     Best
                > >>
                > >>     Julian
                > >>
                > >
                > >
                >




Mime
View raw message