Skip to Content.
Sympa Menu

perfsonar-dev - Re: [pS-dev] Response Message 'type' to unknown Request

Subject: perfsonar development work

List archive

Re: [pS-dev] Response Message 'type' to unknown Request


Chronological Thread 
  • From: Jason Zurawski <>
  • To: Maciej Glowiak <>
  • Cc: Verena Venus <>, "" <>, maxim <>
  • Subject: Re: [pS-dev] Response Message 'type' to unknown Request
  • Date: Mon, 03 Mar 2008 05:04:23 -0500
  • Openpgp: id=B94D59A6; url=http://people.internet2.edu/~zurawski/key.txt
  • Organization: Internet2

All;

>>> This builds upon: https://bugzilla.perfsonar.net/show_bug.cgi?id=331
>>> But
>>> I have not seen the issue brought to the entire list yet.
>>>
>>> In testing messages to services with the wrong MessageType
>>> ('MetadataKeyRequest' is correct, 'MDKeyRequest' is not for example)
>>> the
>>> service must respond with something. I propose that in the event
>>> that a
>>> service does not understand the initial Request it responds with a set
>>> Message response. Currently perfSONAR_PS services will respond with
>>> 'response', although it has been suggested that 'perfSONARResponse',
>>> 'ExceptionResponse' or 'Response' be used instead.
>>>
>>> Since this will effect many services in the upcoming releases I suggest
>>> we implement something quickly so testing can proceed.
>>>
>>> Does anyone have any strong opinions?
>>>
>>
>> At the moment our service responds with MessageType "ErrorReturn" and
>> adds a result code "error.common.action_not_supported" with
>> description "Unknown MessageType: $messagetype".
>>
>> We use "ErrorResponse" in all cases where the parsing of the message
>> content (all below the message tag) is not done to whatever reason.
>>
>> My opinion to this is:
>> The messagetype can be set to two different modes: 1. the proper
>> response type to the request (which btw should be defined somewhere)
>> 2. something else because 1. was not possible
>>
>> In the second case, obviously something utterly wrong happend so this
>> should be indicated by a meaningful name. Just "Response" is not
>> enough if you ask me. I'm not sure if "perfSONARResponse" is right.
>> There might be implementations which use the schemata without
>> perfSONAR, but I'm not sure about this. In any case, I don't see the
>> benefit of naming it "perfSONARResponse". But someone might have good
>> arguments, so please feel free to convince me.
>>
>> My favorites are:
>>
>> "ExceptionResponse" or "ErrorResponse" or somehting along those lines
>> Just my two cents,
>> Verena
>>
>>
>
> Hi, I agree. I like "ErrorResponse" but will accept any other good name.


I think ErrorResponse would fit our general pattern of names thus far.


> Another question - for Release Team is whether we should implement it
> for java services of pS 3.0?
>
> The change won't be complicated, but all developers of Java services
> will need to upgrade ps-base to the newest version supporting this type.
>
> I can do the change very quickly.

We intend to implement this immediately as well, particularly for
perfSONARBOUY and PingER for the MDM release.

-jason



Archive powered by MHonArc 2.6.16.

Top of Page