You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@royale.apache.org by Carlos Rovira <ca...@apache.org> on 2018/06/01 00:36:20 UTC
Re: Royale - BlazeDS working
Reading the document. It uses AMFPHP and Baguetter AMF and the are
transferring 76000 records to a Flex Client. I extracted the most
interesting:
*Adding Baguette AMF plug-in to the basic version of AMF PHP, makes the AMF
faster than JSON and XML when transporting larger amounts of data thus
providing the Baguette AMF. Baguette AMF with AMFPHP is recommended for
enterprise applications which share a large amount of data.*
*On large number of data, AMF is faster than JSON. InFigure 7., on the
basis of measurements in this model, is presented comparison diagram of
dependency between the time and the number of data. Figure 7. shows, that
up to 2000 data JSON is faster than AMF and above this number, AMF becomes
faster. Therefore, it is recommended to use AMF technology above this
number of loaded data.*
*The diagram shows time dependency of the number of data downloaded, with
JSON and XML it is much higher compared to the AMF. The Curve which
describes the AMF technology, the third line, is of considerably smaller
angle, while with JSON, the first line is substantially steeper which means
that the time is increasing faster with the number of loaded data. The
middle line represents the XML format where we can see that the XML much
more depends on JSON when the number of transmitted data increases.*
*In AMF, the period of data transfer does not depend on the amount of data,
because it is binary transmission and is always practically the same.*
*In JSON and XML the connection period depends on the amount of data, the
data is packed in a file in text, and the time of the connection depends on
the size of the file. Therefore, the more data, the connection period
increases with JSON*
*The serialization period on the server’s side depends onthe number of
objects and if there are more objects to be serialized, then more time is
spent which is especially emphasized in JSON and XML because the complexity
is greater. Also the more data there is then the file in JSON and XML are
heavier and more time is needed to transport data*
*Conclusions and Future WorkThe choice of technology depends on several
factors, one of them is the amount of data. By measuring data transfer
period in the authoring application it is determined, for a small amount of
data, up to 2000, JSON technology is recommended because it is faster, and
it is closest to the object structure, orange variant on the graphics. When
large amounts of data are used, up to 2000, AMF technology is recommended,
blue variant on the graphics.*
*When it comes to the complexity of the code, AMF technology is
recommended, JSON and XML in the end*
As you said, our case is different in the deserialization part, but numbers
from output from backend I think are always the same.
We should improve as well algorithms now that we have all this working.
2018-06-01 1:45 GMT+02:00 Carlos Rovira <ca...@apache.org>:
> Is good to know JSON is now more capable, I was imagining that. As you say
> some comparison test will be great. Maybe as we have some time we can do it.
> I assume that many people today will use JSON for new apps if the can
> setup all the interchange with something similar to AMF. If I was to create
> a new system, fro scratch, I'll be doing that test right now to see
> differences and choose JSON or AMF. The great benefit of AMF nowadays,
> without knowing much more about that comparison, is for migration of Flex
> Apps.
>
> 2018-06-01 1:22 GMT+02:00 Alex Harui <ah...@adobe.com>:
>
>> Hi Carlos,
>>
>>
>>
>> I will try to read that in detail later. However, in a quick read, my
>> takeaway is that the findings indicated that for up to 2000 data records,
>> JSON was faster than AMF. And I believe the test was using the Flash
>> Player, where we now have to consider that for Royale, the AMF
>> deserialization is running in JavaScript not in the Flash or browser
>> runtime.
>>
>>
>>
>> Royale does have JSON-to-ValueObject support where the JSON is still
>> parsed in the browser, but objects are instantiated via JavaScript
>> callbacks. And as you mention, I think there are ways to gzip JSON over
>> the wire. So the comparison environment is totally different for Royale vs
>> Flex.
>>
>>
>>
>> So we’ll see if maybe someone can run a test someday and get updated
>> results.
>>
>>
>>
>> Thanks,
>>
>> -Alex
>>
>>
>>
>> *From: *<ca...@gmail.com> on behalf of Carlos Rovira <
>> carlosrovira@apache.org>
>> *Reply-To: *"users@royale.apache.org" <us...@royale.apache.org>
>> *Date: *Thursday, May 31, 2018 at 4:07 PM
>> *To: *"users@royale.apache.org" <us...@royale.apache.org>
>>
>> *Subject: *Re: Royale - BlazeDS working
>>
>>
>>
>> Hi Alex,
>>
>>
>>
>> I saw some comparison document online [1], let me know what do you think
>> about it.
>>
>>
>>
>> For me AMF should not only be measured by its benefits on small size and
>> transmission. In that field I think it should win, since json has more
>> overhead, but as I'm not an expert in JSON, I'm sure some serialization
>> will exists. On that way, maybe JSON, could loose the benefits of the easy
>> format to use that make it popular and will probably be more complex to
>> setup and use, so could make AMF be more attractive. Transferring large
>> amount of data I think AMF will be or at least is what I saw in Census App
>> that James Ward did some years ago. Again, nowadays I'm sure things has
>> evolved and maybe JSON could have some additional complement that make it
>> more tiny but at the cost of more complexity. If someone here is more
>> expert, maybe could give some light on latest JSON advantages. (All this
>> without taking into account additional zlib compress and small messages...)
>>
>>
>>
>> But as I said, for me that's not the main benefit. For me AMF wins in
>> ease of development, since I can code a class in AS3 (a DTO, VO, or POJO,
>> each one can call it in some way) and have the mirror class in the backend
>> (Java, .NET, PHP...). So filling an object in AS3 and sending vía AMF put
>> the object ready to use in the backend method. You don't have to make any
>> process to file some object in the other side, or prepare the data to be
>> used, since this is done under the hood. And that's is very comfortable.
>> Having the same types (the counterpart in the other side) makes all very
>> easy and that's productivity. Maybe JSON can do today something
>> similar...I'm sure something should be out there.
>>
>>
>>
>> Finally for us means to have other tool in our arsenal, a great tool that
>> many people out there doesn't know or thinks on it as it's still some
>> proprietary Adobe technology. And more important, means that many people
>> with Flex apps have now a more easy way to migrate, since they don't need
>> to touch a line of code in their backends. I think that's a huge win. Since
>> instead of planning a full refactor of an entire system (front and back)
>> people can concentrate in front with Royale. Plant side by side with his
>> current Flex App and make it grow from there, and some day turn off flex
>> and turn on Royale :)
>>
>>
>>
>> That's at least my plan with our system :)
>>
>>
>>
>> Thanks!
>>
>>
>>
>> Carlos
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> [1] https://www.researchgate.net/publication/311750440_Comparati
>> ve_analysis_of_AMF_JSON_and_XML_technologies_for_data_
>> transfer_between_the_server_and_the_client
>> <https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.researchgate.net%2Fpublication%2F311750440_Comparative_analysis_of_AMF_JSON_and_XML_technologies_for_data_transfer_between_the_server_and_the_client&data=02%7C01%7Caharui%40adobe.com%7Ca4c6de38ee15443c8f5f08d5c74b5656%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636634048747011993&sdata=U2ljIMDjKcYc6Tiqm%2FI063FksxWtlbMmy0%2FGRgE%2Fzsw%3D&reserved=0>
>>
>>
>>
>> 2018-05-31 21:06 GMT+02:00 Alex Harui <ah...@adobe.com>:
>>
>> Hi Carlos,
>>
>>
>>
>> That’s great news!
>>
>>
>>
>> One thing I would like to see someday (not necessarily from you, maybe
>> some other user) is a comparison of JSON vs AMF. For some real-world
>> server response, what would be the number of bytes transferred and what
>> would be the CPU processing time in the browser to convert the data into
>> “Value Objects”.
>>
>>
>>
>> Thanks,
>>
>> -Alex
>>
>>
>>
>> *From: *<ca...@gmail.com> on behalf of Carlos Rovira <
>> carlosrovira@apache.org>
>> *Reply-To: *"users@royale.apache.org" <us...@royale.apache.org>
>> *Date: *Thursday, May 31, 2018 at 11:52 AM
>> *To: *"users@royale.apache.org" <us...@royale.apache.org>, "
>> dev@royale.apache.org" <de...@royale.apache.org>
>> *Subject: *Re: Royale - BlazeDS working
>>
>>
>>
>> Forgot to share the info in Github page :)
>>
>>
>>
>> https://github.com/apache/royale-asjs/wiki/Apache-Royale-
>> communication-with-AMF-and-RemoteObject
>> <https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fapache%2Froyale-asjs%2Fwiki%2FApache-Royale-communication-with-AMF-and-RemoteObject&data=02%7C01%7Caharui%40adobe.com%7C35acbd7109734fdb4e4d08d5c727a05e%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636633895374721172&sdata=RkbeMMs22sIHVZ0meRLVzGC0VsRUOA1ykl5llHXrvfE%3D&reserved=0>
>>
>>
>>
>> 2018-05-31 20:49 GMT+02:00 Carlos Rovira <ca...@apache.org>:
>>
>> Hi,
>>
>>
>>
>> I get Royale working with BlazeDS. People using BlazeDS should be able to
>> finally start developing with Royale now! :)
>>
>>
>>
>> Right now I'm test it with both the SampleAmfWebApp (in our repo) and
>> with my own backend (real flex/java app), and seems to work in both.
>>
>>
>>
>> Things to have into account: While "clientId" is now working (we send the
>> required PING command, get the DSId and store it for later communication),
>> we need to use it with small messages turned off (blazeds turn it on by
>> default).
>>
>>
>>
>> This is the configuration per channel:
>>
>>
>>
>> <channel-definition ...>
>>
>> <properties>
>>
>> <serialization>
>>
>> <enable-small-messages>false</enable-small-messages>
>>
>> </serialization>
>>
>>
>>
>> I think small messages are only something that will make our
>> implementation shine, but really is not required. I think that will improve
>> a bit over the normal message size. If you use as well the
>> CompressedRemoteObject, that will be a killer communication! :)
>>
>>
>>
>> About small messages, I'm trying to get it work, I uploaded the work
>> currently done, but still needs more to get it fully working.
>>
>>
>>
>> As well I uploaded a SimpleRemoteObject that is a simpler version that
>> should work with AMF backends that require less things like AMFPHP. As I
>> don't use AMFPHP since many years I can say too much about it, but hope
>> others will use and report if something more is needed (The same for other
>> AMF implementations in other technologies).
>>
>>
>>
>> Thanks!
>>
>>
>>
>> --
>>
>> Carlos Rovira
>>
>> http://about.me/carlosrovira
>> <https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fabout.me%2Fcarlosrovira&data=02%7C01%7Caharui%40adobe.com%7C35acbd7109734fdb4e4d08d5c727a05e%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636633895374721172&sdata=%2F%2FeMiAc533chEQ5po1rV%2B9tMGMLEja4Yty%2BMg0t1j8U%3D&reserved=0>
>>
>>
>>
>>
>>
>>
>>
>> --
>>
>> Carlos Rovira
>>
>> http://about.me/carlosrovira
>> <https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fabout.me%2Fcarlosrovira&data=02%7C01%7Caharui%40adobe.com%7C35acbd7109734fdb4e4d08d5c727a05e%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636633895374877422&sdata=DJRVjcGCk26lBJoBify1%2B9O0TDqRZEGdNu4ARqglWLA%3D&reserved=0>
>>
>>
>>
>>
>>
>>
>>
>> --
>>
>> Carlos Rovira
>>
>> http://about.me/carlosrovira
>> <https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fabout.me%2Fcarlosrovira&data=02%7C01%7Caharui%40adobe.com%7Ca4c6de38ee15443c8f5f08d5c74b5656%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636634048747032007&sdata=7RXif5%2FxYB0HvseC2hvhehkKflRT8Obi4C7Qr27hPDY%3D&reserved=0>
>>
>>
>>
>
>
>
> --
> Carlos Rovira
> http://about.me/carlosrovira
>
>
--
Carlos Rovira
http://about.me/carlosrovira
Re: Royale - BlazeDS working
Posted by Carlos Rovira <ca...@apache.org>.
Hi Alex,
Justing doesn't refer to James's implementation of amf.js he did around
2010, but to the old Flex Census app, that was a Flex app that runs a
series of tests.
I think it could be of help to make our own. But is clear, as you said,
that to get our numbers, we need to create our own from scratch.
2018-06-01 18:46 GMT+02:00 Alex Harui <ah...@adobe.com>:
> We are not using James Ward's code, so any findings he had would be of
> questionable use to us. Even tests against the amf.js we did start from
> are potentially invalid because we plugged our version into a ClassAlias
> registry. It is best if we actually test our code and not somebody else's.
>
> My 2 cents,
> -Alex
>
> On 6/1/18, 12:56 AM, "Justin Mclean" <ju...@classsoftware.com> wrote:
>
> Hi,
>
> > thanks for the link. That's the one I was referring ni previous
> emails. Since is about 10 years old, the old census app seems not to be
> online. At that time AMF improvements over JSON was huge. I think those
> days will still be great, but maybe the gap is smaller.
>
> You might want to search the archives (or internet way back machine) I
> have a vague recollection of James making the code to that app available.
>
> Thanks,
> Justin
>
>
--
Carlos Rovira
http://about.me/carlosrovira
Re: Royale - BlazeDS working
Posted by Alex Harui <ah...@adobe.com>.
We are not using James Ward's code, so any findings he had would be of questionable use to us. Even tests against the amf.js we did start from are potentially invalid because we plugged our version into a ClassAlias registry. It is best if we actually test our code and not somebody else's.
My 2 cents,
-Alex
On 6/1/18, 12:56 AM, "Justin Mclean" <ju...@classsoftware.com> wrote:
Hi,
> thanks for the link. That's the one I was referring ni previous emails. Since is about 10 years old, the old census app seems not to be online. At that time AMF improvements over JSON was huge. I think those days will still be great, but maybe the gap is smaller.
You might want to search the archives (or internet way back machine) I have a vague recollection of James making the code to that app available.
Thanks,
Justin
Re: Royale - BlazeDS working
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
> thanks for the link. That's the one I was referring ni previous emails. Since is about 10 years old, the old census app seems not to be online. At that time AMF improvements over JSON was huge. I think those days will still be great, but maybe the gap is smaller.
You might want to search the archives (or internet way back machine) I have a vague recollection of James making the code to that app available.
Thanks,
Justin
Re: Royale - BlazeDS working
Posted by Carlos Rovira <ca...@apache.org>.
Hi Justin,
thanks for the link. That's the one I was referring ni previous emails.
Since is about 10 years old, the old census app seems not to be online. At
that time AMF improvements over JSON was huge. I think those days will
still be great, but maybe the gap is smaller.
Thanks!
Carlos
2018-06-01 3:43 GMT+02:00 Justin Mclean <ju...@classsoftware.com>:
> Hi,
>
> You might want to look at this for some performance data as well. It
> probably could be repurposed and create more current bench marks?
>
> https://www.jamesward.com/2007/04/30/ajax-and-flex-data-
> loading-benchmarks/
>
> Thanks,
> Justin
--
Carlos Rovira
http://about.me/carlosrovira
Re: Royale - BlazeDS working
Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,
You might want to look at this for some performance data as well. It probably could be repurposed and create more current bench marks?
https://www.jamesward.com/2007/04/30/ajax-and-flex-data-loading-benchmarks/
Thanks,
Justin