Hi all, just some feedback on my efforts towards formal definition of protocol message types:
It seems that DFDL is a perfect match for defining the data formats. Right now I'm working hard on writing a DFDL schema for the S7 Protocol as this has quite a number of "specialities" that provide a challenge when writing such a definition. As soon as that's finished I guess we know if this is the way to go. Unfortunately this is where the real work begins. As Daffodil (The Apache implementation of a DFDL framework ... the other one is from IBM but not so feature-rich) parses any data using DFDL schemas into a JSON or XML presentation. What it currently doesn't do is parse into a generic model and/or generate code. This is where I will probably have to invest some time in the Apache Daffodil Incubator podling and help them implement the missing parts. I know this is quite some work, but it will be a HUGE benefit for our project and it will also help our brother project greatly. What will be possible with Daffodil, is that we can replace the "pcapng" binary captures with XML (or JSON) versions which Daffodil can already serialize to binary data for tests ... So we get human readable test-case input. We can dump incoming packets into a human readable form and eventually (automatically) generate test-case input for inspecting IO problems. Also could we implement a low-performance version of new protocols, by using the existing Daffodil parsers/serializers to process a protocol specification to and from XML/JSON ... this should be a huge benefit for new-protocol-implementors. As soon as a protocol is roughly implemented, we can switch to generated model, (de)serializers. So far the update and a little excuse for my silence ;-) Chris Am 10.01.19, 10:41 schrieb "Christofer Dutz" <[email protected]>: Hi Markus, that is not quite true. In some cases definitely yes, but I'm currently looking into the option of using DFDL (Apache Daffodil) for this. The way I would like to do it, would be to use DFDL to describe the format and then generate code from that with a general purpose generator for which we can provide the code-generation templates. This way we would have to write these templates once and exactly for PLC4X and then have all protocols generated to perfectly fit. Right now I am not doing anything different ... I'm trying to refactor things in a way that is as generic as possible without any performance drawbacks and then to adjust one protocol after the other to match that. I would like to automate this. So as soon as a new language should be supported, someone would setup the general skeleton driver, write a prototype and then convert that into templates and "presto" we get all the layers generated. At least that's what I'm currently thinking of. Chris Am 10.01.19, 10:33 schrieb "Markus Sommer" <[email protected]>: Hello everyone, If we go the way of automatic object generation for messages, then we will have to live with performance degradation. Best regards Markus Mit freundlichen Grüßen Markus Sommer Geschäftsführer isb innovative software businesses GmbH Otto-Lilienthal-Straße 2 D 88046 Friedrichshafen Tel +49 (0) 7541 3834-14 Fax +49 (0) 7541 3834-20 Mobil +49 (0) 171 537 8437 Mail to [email protected] http://www.isb-fn.de Geschäftsführer: Markus Sommer, Thomas Zeler Registergericht: Amtsgericht Ulm HRB-Nr. 631624 Important Note: This e-mail and any attachments are confidential, may contain trade secrets and may well also be legally privileged or otherwise protected from disclosure. If you have received it in error, you are on notice of its status. Please notify us immediately by reply e-mail and then delete this e-mail and any attachment from your system. If you are not the intended recipient please understand that you must not copy this e-mail or any attachments or disclose the contents to any other person. Thank you. -----Ursprüngliche Nachricht----- Von: Christofer Dutz <[email protected]> Gesendet: Mittwoch, 9. Januar 2019 18:59 An: [email protected] Betreff: Re: Antwort: Re: Antwort: Re: Antwort: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? Well I thought that went could use what's already there as I was expecting us but to be the first. But I haven't found an option that works. Of course whipping up a coffee generator is quite simple (done it loads of times) but you have to come up with the serialisation for all supported languages ourself. I wanted to avoid that, cause it's a lot of work. I think we should do a little more evaluating. If we don't find anything well go down that path. But having plugins used in a build that are also part of the build itself introduces more issues. So perhaps becoming involved in the daffodil project and adding a code generator there is the better option. (Code generator in Daffodil and generation templates here) Chris Outlook for Android<https://aka.ms/ghei36> herunterladen ________________________________ From: Julian Feinauer <[email protected]> Sent: Wednesday, January 9, 2019 6:46:14 PM To: [email protected] Subject: Re: Antwort: Re: Antwort: Re: Antwort: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? Okay, I see... When I learned something, that tooling is really important, and it makes me a bit of sorrows that DFDL is not there yet. Do I take things too simple if we would just starting using something like json (or xml, as I think chris likes xml better) to define these messages and then implement a (very simple) converter to the class format? Shouldn’t this be doable? Of course there's still the lack of a Maven Plugin but I think this could be made easily (????) if its just invocing a method (or we do something like ant task voodoo to invoke it via CMD). But at least we would have nice syntax highlighting and such stuff and could rely on Jackson to get the files to a Java Input (and in case of xml we could also do validation directly and generate nice HTML documentatsions via xlst). Am I missing something or take things too simple? Best Julian Am 09.01.19, 18:26 schrieb "Christofer Dutz" <[email protected]>: Hi Julian, Yes exactly ... Unfortunately the messages are composed of different types of structures. Header, parameters of different type and length, payloads of different type. But in general, yes Chris Outlook for Android<https://aka.ms/ghei36> herunterladen ________________________________ From: Julian Feinauer <[email protected]> Sent: Wednesday, January 9, 2019 6:16:01 PM To: [email protected] Subject: Re: Antwort: Re: Antwort: Re: Antwort: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? Hi Chris, I'm not sure if I got it right, but what we would need is a way to generate classes in multiple languages based on a message definition, or? So we say something like Bit, byte, byte, bit And it generates a class (or struct?) which deserializes / serializes to that, or? Sorry for that (probably) dump question but I got a bit confused by this protobuf / thrift approach. Julian Am 09.01.19, 17:49 schrieb "Christofer Dutz" <[email protected]>: Hi Sebastian, Well in general the difference between a read and a write in S7 is one byte value ;-) The overall structure is somewhat identical. Chris Von: Sebastian Wiendl <[email protected]> Antworten an: "[email protected]" <[email protected]> Datum: Mittwoch, 9. Januar 2019 um 17:21 An: "[email protected]" <[email protected]> Betreff: Antwort: Re: Antwort: Re: Antwort: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? The developer's struggle: Which foot to shoot yourself in. ;-) I have no detailed knowledge about the protocols you want to serialize/deserialize - maybe they are simple enough that the mentioned early serialization features of Kaitai are sufficient. If not, I wouldn't recommend using it. Another aspect might be the anticipated distribution of read and write workloads of a typical PLC4X application - if its mostly reads it might still be worth a try, because you save a lot of (implementation) effort on the deserialization which frees up resources for serialization or other things in general. Mit freundlichen Grüßen Kind regards Sebastian Wiendl DSE / Digital Solutions Software Engineer Phone: +49 9605 919 - 9341 E-Mail: [email protected] Internet: http://www.bhs-world.com ________________________________ [cid:_1_13C7BE3013C7BA340059D4ACC125837D] BHS Corrugated Maschinen- und Anlagenbau GmbH Paul-Engel-Straße 1 92729 WEIHERHAMMER GERMANY Management: Christian Engel, Lars Engel, Norbert Städele Registered at Amtsgericht Weiden, HR B 1320 [cid:_2_13C7CB4413C7C7340059D4ACC125837D] ________________________________ Diese Nachricht ist nur für den Empfänger bestimmt, da sie persönliche und/oder rechtlich geschützte Informationen und vertrauliche Informationen enthält. Der Inhalt darf an keinen anderen außer an den Empfänger weitergeleitet werden. Unbefugte Empfänger werden gebeten, diese Vertraulichkeit zu wahren und den Absender sofort von diesem Übertragungsfehler in Kenntnis zu setzen. Vernichten Sie diese Nachricht. Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Nachricht ist nicht gestattet. This message is intended for the addressee only as it contains private and/or privileged and confidential information. The contents are not to be disclosed to anyone else than the addressee. Unauthorized recipients are requested to comply with the above and to inform the sender immediately of any errors in transmission and destroy this e-mail. Any unauthorized copying, disclosure or distribution of the material in this e-mail is strictly prohibited. Von: "Christofer Dutz" <[email protected]> An: "[email protected]" <[email protected]> Datum: 09.01.2019 17:09 Betreff: Re: Antwort: Re: Antwort: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? ________________________________ Hi Sebastian, yeah … I asked in the Incubator list and they told me it would be safe to use, if the output was not GPLed … so I contacted the maintainer of the project and simply asked him. He told me that the output would match the input license. So if our definitions would be Apache 2.0, so would the output … so it seems we would be safe and able to use that. However he also told me this: “That's great news! Please note, however, that serialization is in its early stages in Kaitai Struct, so it might be no exactly up to your expectations.“ So I don’t know if we should go down that path … right now … :-/ Chris Von: Sebastian Wiendl <[email protected]> Antworten an: "[email protected]" <[email protected]> Datum: Mittwoch, 9. Januar 2019 um 16:46 An: "[email protected]" <[email protected]> Betreff: Antwort: Re: Antwort: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? Another thing to consider: Last time I checked Kaitai only supported deserialization (https://github.com/kaitai-io/kaitai_struct/issues/27). Might be inconvenient to introduce another framework for serialization... Mit freundlichen Grüßen Kind regards Sebastian Wiendl DSE / Digital Solutions Software Engineer Phone: +49 9605 919 - 9341 E-Mail: [email protected] Internet: http://www.bhs-world.com<http://www.bhs-world.com/> ________________________________ [cid:_1_147DDC68147DD8840056A390C125837D] BHS Corrugated Maschinen- und Anlagenbau GmbH Paul-Engel-Straße 1 92729 WEIHERHAMMER GERMANY Management: Christian Engel, Lars Engel, Norbert Städele Registered at Amtsgericht Weiden, HR B 1320 [cid:_2_147DE97C147DE56C0056A390C125837D] ________________________________ Diese Nachricht ist nur für den Empfänger bestimmt, da sie persönliche und/oder rechtlich geschützte Informationen und vertrauliche Informationen enthält. Der Inhalt darf an keinen anderen außer an den Empfänger weitergeleitet werden. Unbefugte Empfänger werden gebeten, diese Vertraulichkeit zu wahren und den Absender sofort von diesem Übertragungsfehler in Kenntnis zu setzen. Vernichten Sie diese Nachricht. Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Nachricht ist nicht gestattet. This message is intended for the addressee only as it contains private and/or privileged and confidential information. The contents are not to be disclosed to anyone else than the addressee. Unauthorized recipients are requested to comply with the above and to inform the sender immediately of any errors in transmission and destroy this e-mail. Any unauthorized copying, disclosure or distribution of the material in this e-mail is strictly prohibited. Von: "Christofer Dutz" <[email protected]> An: "[email protected]" <[email protected]> Datum: 09.01.2019 15:57 Betreff: Re: Antwort: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? ________________________________ The thing is that we would actually not be bundling that … the compiler is only used during the build and not at runtime. I am currently double-checking this with the others in the incubator mailing-list. So it’s not completely impossible … let’s see what the others have to say about it. But I agree … it does look as if it would suit our needs. Chris Von: Sebastian Wiendl <[email protected]> Antworten an: "[email protected]" <[email protected]> Datum: Mittwoch, 9. Januar 2019 um 15:42 An: "[email protected]" <[email protected]> Betreff: Antwort: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? Yeah, i just checked... bummer the compiler is GPLed... The project itself has some really good ideas how to handle binary parsing. I used it prototypically in decoding a proprietary UDP message format and it worked great. Mit freundlichen Grüßen Kind regards Sebastian Wiendl DSE / Digital Solutions Software Engineer Phone: +49 9605 919 - 9341 E-Mail: [email protected] Internet: http://www.bhs-world.com<http://www.bhs-world.com/><http://www.bhs-world.com/> ________________________________ [cid:_1_147D3C84147D38880050BA3BC125837D] BHS Corrugated Maschinen- und Anlagenbau GmbH Paul-Engel-Straße 1 92729 WEIHERHAMMER GERMANY Management: Christian Engel, Lars Engel, Norbert Städele Registered at Amtsgericht Weiden, HR B 1320 [cid:_2_147D4998147D45880050BA3BC125837D] ________________________________ Diese Nachricht ist nur für den Empfänger bestimmt, da sie persönliche und/oder rechtlich geschützte Informationen und vertrauliche Informationen enthält. Der Inhalt darf an keinen anderen außer an den Empfänger weitergeleitet werden. Unbefugte Empfänger werden gebeten, diese Vertraulichkeit zu wahren und den Absender sofort von diesem Übertragungsfehler in Kenntnis zu setzen. Vernichten Sie diese Nachricht. Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Nachricht ist nicht gestattet. This message is intended for the addressee only as it contains private and/or privileged and confidential information. The contents are not to be disclosed to anyone else than the addressee. Unauthorized recipients are requested to comply with the above and to inform the sender immediately of any errors in transmission and destroy this e-mail. Any unauthorized copying, disclosure or distribution of the material in this e-mail is strictly prohibited. Von: "Christofer Dutz" <[email protected]> An: "[email protected]" <[email protected]> Datum: 09.01.2019 15:37 Betreff: Re: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? ________________________________ Hi Sebasitan, thanks for that pointer … unfortunately I think we might be unable to use this option as the compiler is GPLv3 licensed … that is a category X license. Even if the compiler is not bundled with our software or used at runtime (will probably only need it at compile-time), I doubt we would be allowed to use it. But still I’ll look into it … Chris Von: Sebastian Wiendl <[email protected]> Antworten an: "[email protected]" <[email protected]> Datum: Mittwoch, 9. Januar 2019 um 15:18 An: "[email protected]" <[email protected]> Betreff: Antwort: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? Hi, maybe the Kaitai Struct project (https://kaitai.io/) can help you achieve this? Mit freundlichen Grüßen Kind regards Sebastian Wiendl DSE / Digital Solutions Software Engineer Phone: +49 9605 919 - 9341 E-Mail: [email protected] Internet: http://www.bhs-world.com<http://www.bhs-world.com/><http://www.bhs-world.com/><http://www.bhs-world.com/> ________________________________ [cid:_1_0DB11BC80DB117CC004E9C08C125837D] BHS Corrugated Maschinen- und Anlagenbau GmbH Paul-Engel-Straße 1 92729 WEIHERHAMMER GERMANY Management: Christian Engel, Lars Engel, Norbert Städele Registered at Amtsgericht Weiden, HR B 1320 [cid:_2_0DB128DC0DB124CC004E9C08C125837D] ________________________________ Diese Nachricht ist nur für den Empfänger bestimmt, da sie persönliche und/oder rechtlich geschützte Informationen und vertrauliche Informationen enthält. Der Inhalt darf an keinen anderen außer an den Empfänger weitergeleitet werden. Unbefugte Empfänger werden gebeten, diese Vertraulichkeit zu wahren und den Absender sofort von diesem Übertragungsfehler in Kenntnis zu setzen. Vernichten Sie diese Nachricht. Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Nachricht ist nicht gestattet. This message is intended for the addressee only as it contains private and/or privileged and confidential information. The contents are not to be disclosed to anyone else than the addressee. Unauthorized recipients are requested to comply with the above and to inform the sender immediately of any errors in transmission and destroy this e-mail. Any unauthorized copying, disclosure or distribution of the material in this e-mail is strictly prohibited. Von: "Christofer Dutz" <[email protected]> An: "[email protected]" <[email protected]> Datum: 09.01.2019 15:06 Betreff: Re: Anyone got experience with "protocol buffers" or DFDL (for generation of the message (de)serialization code)? ________________________________ Hi all, Ok ... so protobuf seems to be semi-ideal ... It seems that you can use it to model the structure of data. Protobuf is good for generating model classes, parsers and serializers for a given model ... the binary data-format is a result of this. We want the opposite: We want to generate a model from a known output data-format. In general this could be somehow achieved with protobuf, however it is very difficult to produce the definition in a way that it is able to parse a given data format. For example simply outputting one byte seems to be problematic. I was able to somehow hack an enum and provide some extension to allow providing code values, but we don't have the level of control we would need to and the result is not very readable. I was able to quite easily setup the maven build to generate java code for parsing and serializing a model ... so that was good. DFDL looks as if it's ideal for describing the data format, however I couldn't find tooling to generate model, parser and serializer from a DFDL definition. I subscribed to our brother incubating project Daffodil and asked on their list ... perhaps I have to get my hands dirty and implement the maven plugin and code generators as part of that project ... I am hoping not having to do that. I'll check out Thrift in parallel ;-) Chris Am 09.01.19, 11:19 schrieb "Christofer Dutz" <[email protected]>: From my first look at thrift some time ago, that's more about API and not about the actual payload, is it? How about I try to do a protobuf version of the "s7-protocol" and you give thrift a try? Another option would be the DFDL option. Chris Am 09.01.19, 11:13 schrieb "Julian Feinauer" <[email protected]>: Hi Chris, we worked (and work) with Thrift [1] at several places. Thrift is a strong contender to protobuf and both have their specific advantages and disadvantages. Perhaps I would prefer Thrift as it comes from the Apache Ecosystm (and supports more langauges) but generally, Tim can say more about working with Thrift. Best Julian [1] https://thrift.apache.org/ Am 09.01.19, 10:45 schrieb "Christofer Dutz" <[email protected]>: Hi all, while I’m currently working on refactoring the S7 driver to a simpler structure so we can convert it to other languages more easily. A colleague of mine pointed me to protobuf/protocol buffers from google [1] From a quick look at it, it does seem as if it could suit our needs quite nicely. I would like to try out if it’s possible to model the S7 data structures in this way. If it works we could eventually quickly create something that serializes/deserializes given data in any language … It seems to be a lot simpler than the DFDL [2] I was thinking of, so guess we have to find out if it has all the capabilities we need. Any thoughts? Chris [1] https://developers.google.com/protocol-buffers/docs/javatutorial [2] https://en.wikipedia.org/wiki/Data_Format_Description_Language
