Home » railML newsgroups » railml.timetable » constraints for OperatingPeriod
constraints for OperatingPeriod [message #813] Tue, 25 September 2012 15:00 Go to next message
Andreas Tanner is currently offline  Andreas Tanner
Messages: 52
Registered: March 2012
Member
The operatingPeriod type currently is modelled quite flexibly to
accommodate different use cases. However, the standard does not define
which combinations of attributes can meaningfully be used together. A
stricter definition would spare a a lot of discussions between different
users of the standard. Here is a suggestion:

***
The OperatingPeriod element can be used for three different use cases.

- the calendar based operating period:
-- bitmask and startDate are mandatory,
-- endDate, operatingDay, specialService are not allowed. [endDate is
not allowed since it would be redundant with startDate + bitmask length]

- standard week operating period:
-- operatingDay is mandatory (at least one),
-- specialService optional
-- startDate, endDate are optional and if used, both must be given
-- bitmask is not allowed

- abstract operating period
-- name or code are mandatory
-- bitmask, operatingDay, specialService are not allowed.

Always allowed, and optional if not declared otherwise, are name,
additionalName, code, description, timetablePeriodRef, xml:lang, dayOffset

***

For railML 3.0, I would suggest to model the three cases as distinct
types CalendarBasedOperatingPeriod, etc, all derived from base class
OperatingPeriod.

Best regards
--Andreas Tanner.
Re: constraints for OperatingPeriod [message #815 is a reply to message #813] Tue, 02 October 2012 17:34 Go to previous messageGo to next message
Dirk Bräuer is currently offline  Dirk Bräuer
Messages: 311
Registered: August 2008
Senior Member
Dear Andreas,

when the current operatingPeriod type was defined, the aim was to
explicitly allow a 'describing' definition and a bitMask redundantly. The
background is that there are many 'describing' expressions for the same
bitmask. For instance: "Mo-Fr, 01.08-31.08 only" and "[Sa+So], not until
31.07 and not from 01.09." lead to the same bitmask.

In some cases,
- we do want to have the bitmask for easier and clearer reading of the
real operating days,
- we do want to have the 'describing' expressions to exactly reconstruct
the input of the user, meaning to 'show' the contents of the bitmask to
the user in a kind which is familiar for him and which reflects the input.
If the user did input "Mo-Fr[S]", he should not be shown "W[Sa]" and vice
versa.

Please have a look at the examples in
www.irfp.de/download/railml_beispiel_verkehrstage.pdf

This means that
> - the calendar based operating period:
> -- bitmask and startDate are mandatory,
> -- endDate, operatingDay, specialService are not allowed.
would not fit to the current requirements.

---
Concerning a 'third' combination
> - abstract operating period
> -- name or code are mandatory
> -- bitmask, operatingDay, specialService are not allowed.

I would welcome such a possibility. But, from my side we need a definition
whether different 'abstract operating periods' are disjunctive or not.
Either we define a matrix or, at least, we define that different 'abstract
operating periods' _always_ have to be disjunctive.

Additionally, I would prefer to allow an abstract operating period to
refer to a 'real' operating period. In my opinion, any abstract operating
period earlier or later becomes real. So, we can have a timetable which
was planned with abstract operating periods (in the first instance) but in
the meantime, the user does know which real dates are assigned to which
abstract operating period.

So, an abstract operating period can have
- either a bitmask (at least)
- or a reference to another (non-abstract) operating period.

From these suggestions, I would prefer
- 'abstract operating periods' always have to be disjunctive
- 'abstract operating periods' can optionally reference another
(non-abstract) operating period.

Best regards,
Dirk.
Re: constraints for OperatingPeriod [message #823 is a reply to message #815] Mon, 15 October 2012 09:35 Go to previous messageGo to next message
Andreas Tanner is currently offline  Andreas Tanner
Messages: 52
Registered: March 2012
Member
Am 02.10.2012 17:34, schrieb Dirk Bräuer:
> Dear Andreas,
>
> when the current operatingPeriod type was defined, the aim was to
> explicitly allow a 'describing' definition and a bitMask redundantly.
> The background is that there are many 'describing' expressions for the
> same bitmask. For instance: "Mo-Fr, 01.08-31.08 only" and "[Sa+So], not
> until 31.07 and not from 01.09." lead to the same bitmask.
>

This is two different issues:
1) Redundancy
2) "describing" expressions.
With the latter, I agree, but I intended them to be covered in the
"standard week" case. It does include the specialService etc.
constructions. It should be named more appropriately, though, maybe
"rule based" or something like that.

With the redundancy, I do have a problem. It does allow to be lenient
when /writing/ railML, but the costs incur at the import side. For
serious import software, one has to
- extend the customer-specific specification as to disallow
inconsistencies between bitmask and rule
- code a check of the resulting precondition
- add a test case for the software.

So please, let us abstain from casually sprinkle some redundancy through
the standard for "easer and clearer reading".

>
> ---
> Concerning a 'third' combination
>> - abstract operating period
>> -- name or code are mandatory
>> -- bitmask, operatingDay, specialService are not allowed.
>
> I would welcome such a possibility. But, from my side we need a
> definition whether different 'abstract operating periods' are
> disjunctive or not. Either we define a matrix or, at least, we define
> that different 'abstract operating periods' _always_ have to be
> disjunctive.

This is an important point. In
http://www.irfp.de/download/railml_doku_beispiele.pdf, you write on the
issue of /keys/ for a train. In another thread in this forum, delay
infos are discussed. There, for instance, one may want to state that
"that instance of the 12345 train with operating period "mo-fr except
holiday" that runs on oct 13, is delayed". So we need even more than an
understanding just of disjoint operating period, namely, a concept of
unrolling rule-based operating periods onto a calendar, and identify
"instances" within a period (which is easy once we have bitmasks
/resulting from unrolling/).


>
> Additionally, I would prefer to allow an abstract operating period to
> refer to a 'real' operating period. In my opinion, any abstract
> operating period earlier or later becomes real.

Here I understand that you aim at the mapping of different stages in the
planning process. I agree that this is important. I think that this
topic would need more analysis.

Maybe there could be a slot at the upcoming railML conference to discuss
these issues?

Best, --Andreas.
Re: constraints for OperatingPeriod [message #825 is a reply to message #823] Wed, 17 October 2012 20:11 Go to previous messageGo to next message
Dirk Bräuer is currently offline  Dirk Bräuer
Messages: 311
Registered: August 2008
Senior Member
Dear Andreas,

> With the redundancy, I do have a problem. It does allow to be lenient
> when /writing/ railML, but the costs incur at the import side. For
> serious import software, one has to
> - extend the customer-specific specification as to disallow
> inconsistencies between bitmask and rule
> - code a check of the resulting precondition
> - add a test case for the software.

I am aware the problems of redundancies when importing. I do not at all
defend them in general.

You do not need to "disallow inconsistencies between bitmask and rule"
when importing. You can describe which elements and attributes of RailML
are relevant for your software and which are not. (From my opinion, you
have to create an own Interface Documentation for a serious import
software.) There you can consider either the bit mask or the rule to be
relevant and to be exclusive alternatives for your software. If you feel
necessary to close all possible inconsistencies you can do, but this is
not requested by RailML and probably not completely possible. (For
instance, when importing a timetable, you will probably not import all
infrastructure and all vehicle data in detail just to check whether there
are inconsistencies in it...)

Please consider:
When writing a RailML file, the software does normally not know for which
purpose the RailML file will be used. It has to create a RailML file which
is most possibly general.
When reading a RailML file, the software can exactly know the
requirements of the target system and therefore can decide which elements
and attributes are relevant and which additional rules apply. From my
opinion, there always will be additional (semantical) rules which are out
of the scope of RailML.

RailML does not implicitly avoid all problems that may occur. Rather,
RailML gives you a basis not to "reinvent the wheel" again and again each
time you need an interface to/from another software.

> So please, let us abstain from casually sprinkle some redundancy through
> the standard for "easer and clearer reading".

Well, ok, but in practice we have to be aware that
- if we always implement the "most scientific, exact but most complicated
way" to describe a case only, to totally avoid redundancies, we risk that
RailML will not be accepted by other software developers because of being
too much difficult. It should not be our aim to create a RailML which is
free from redundancies but nobody will use.
- sometimes we have to implement a 'short' solution possibly before we
ourselves are aware the more general solution. Then, later, when needing
the more general solution, we cannot delete the 'short' solution
immediately because of the rules of downward compatibility. So, we have a
redundancy at least temporarily.

> a concept of unrolling rule-based operating periods onto a calendar,

I agree. This should have been the bit mask.

> and identify "instances" within a period (which is easy once we have
> bitmasks /resulting from unrolling/).

I have some problem with these 'actual' (running) information in a schema
called 'timetable'. I think the term 'timetable' implements that it is
about planning only, not about 'actual' (running) information.

But, this should not mean that such information should not be possible in
RailML. I am not sure whether it is possible to describe the aspects of
'actual' (running) information in a proper way with current RailML. We
have some information about that but they are more dead bodies from
RailML1...

Anyway, I agree with you: We would need a possibility to identify
"instances" within a period to describe 'actual' information additionally
to 'timetable' information - either in 'timetable' schema or somewhere
else.

>> Additionally, I would prefer to allow an abstract operating period to
>> refer to a 'real' operating period. In my opinion, any abstract
>> operating period earlier or later becomes real.
>
> Here I understand that you aim at the mapping of different stages in the
> planning process.

Yes. I exactly "aimed" on

> a concept of unrolling rule-based operating periods onto a calendar

So far, this should have been the bit mask. But again: RailML is not
perfect nor complete...

Best regards,
Dirk.
Re: constraints for OperatingPeriod [message #861 is a reply to message #825] Fri, 09 November 2012 10:27 Go to previous messageGo to next message
Susanne Wunsch railML is currently offline  Susanne Wunsch railML
Messages: 0
Registered: January 2020
Hi Dirk, Andreas and others,

Dirk Bräuer <dirkbraeuer(at)irfpde> writes:
>> With the redundancy, I do have a problem. It does allow to be
>> lenient when /writing/ railML, but the costs incur at the import
>> side. For serious import software, one has to
>> - extend the customer-specific specification as to disallow
>> inconsistencies between bitmask and rule
>> - code a check of the resulting precondition
>> - add a test case for the software.
>
> Please consider:
> When writing a RailML file, the software does normally not know for
> which purpose the RailML file will be used. It has to create a RailML
> file which is most possibly general.
> When reading a RailML file, the software can exactly know the
> requirements of the target system and therefore can decide which
> elements and attributes are relevant and which additional rules
> apply. From my opinion, there always will be additional (semantical)
> rules which are out of the scope of RailML.

Thank you, Dirk, for the above described explanations.

Andreas, are you convinced by them?

How about the wanted re-structuring of 'operatingPeriod' for the next
major release? Can we drop it? Or do you propose another easier to
implement/validate/understand structure?

> Anyway, I agree with you: We would need a possibility to identify
> "instances" within a period to describe 'actual' information
> additionally to 'timetable' information - either in 'timetable'
> schema or somewhere else.
>
>>> Additionally, I would prefer to allow an abstract operating period to
>>> refer to a 'real' operating period. In my opinion, any abstract
>>> operating period earlier or later becomes real.
>>
>> Here I understand that you aim at the mapping of different stages in
>> the planning process.
>
> Yes. I exactly "aimed" on
>
>> a concept of unrolling rule-based operating periods onto a calendar
>
> So far, this should have been the bit mask.

I filed a ticket for the issue of an "abstract operating period":

http://trac.assembla.com/railML/ticket/187

Kind regards...
Susanne

--
Susanne Wunsch
Schema Coordinator: railML.common
Re: constraints for OperatingPeriod [message #864 is a reply to message #861] Fri, 09 November 2012 10:50 Go to previous messageGo to next message
Andreas Tanner is currently offline  Andreas Tanner
Messages: 52
Registered: March 2012
Member
Dear all,
as far as railML 2.x is concerned, my suggestion was just to enhance the
documentation. The discussion on redundancy has some philosophical
traits and therefore, absolute truth cannot be expected to be found.

For railML 3.0, however, I would like to keep the discussion open and
let us elaborate a model of validity with a more formally defined
semantics that allows to address specific instances of trains from a
timetable.

Best, Andreas.

Am 09.11.2012 10:27, schrieb Susanne Wunsch:
> Hi Dirk, Andreas and others,
>
> Dirk Bräuer <dirkbraeuer(at)irfpde> writes:
>>> With the redundancy, I do have a problem. It does allow to be
>>> lenient when /writing/ railML, but the costs incur at the import
>>> side. For serious import software, one has to
>>> - extend the customer-specific specification as to disallow
>>> inconsistencies between bitmask and rule
>>> - code a check of the resulting precondition
>>> - add a test case for the software.
>>
>> Please consider:
>> When writing a RailML file, the software does normally not know for
>> which purpose the RailML file will be used. It has to create a RailML
>> file which is most possibly general.
>> When reading a RailML file, the software can exactly know the
>> requirements of the target system and therefore can decide which
>> elements and attributes are relevant and which additional rules
>> apply. From my opinion, there always will be additional (semantical)
>> rules which are out of the scope of RailML.
>
> Thank you, Dirk, for the above described explanations.
>
> Andreas, are you convinced by them?
>
> How about the wanted re-structuring of 'operatingPeriod' for the next
> major release? Can we drop it? Or do you propose another easier to
> implement/validate/understand structure?
>
>> Anyway, I agree with you: We would need a possibility to identify
>> "instances" within a period to describe 'actual' information
>> additionally to 'timetable' information - either in 'timetable'
>> schema or somewhere else.
>>
>>>> Additionally, I would prefer to allow an abstract operating period to
>>>> refer to a 'real' operating period. In my opinion, any abstract
>>>> operating period earlier or later becomes real.
>>>
>>> Here I understand that you aim at the mapping of different stages in
>>> the planning process.
>>
>> Yes. I exactly "aimed" on
>>
>>> a concept of unrolling rule-based operating periods onto a calendar
>>
>> So far, this should have been the bit mask.
>
> I filed a ticket for the issue of an "abstract operating period":
>
> http://trac.assembla.com/railML/ticket/187
>
> Kind regards...
> Susanne
>
Re: constraints for OperatingPeriod [message #871 is a reply to message #864] Fri, 09 November 2012 15:51 Go to previous messageGo to next message
Susanne Wunsch railML is currently offline  Susanne Wunsch railML
Messages: 0
Registered: January 2020
Dear Andreas and others,

Andreas Tanner <ata(at)ivude> writes:

> as far as railML 2.x is concerned, my suggestion was just to enhance
> the documentation.

Would you be so kind as to enhance the wiki documentation by the
findings of this thread yourselves? If you need any assistance do not
hesitate to ask for.

http://www.wiki.railml.org/index.php?title=TT:operatingPerio d

> The discussion on redundancy has some philosophical traits and
> therefore, absolute truth cannot be expected to be found.

That's something we should be aware of every time we develop railML.

> For railML 3.0, however, I would like to keep the discussion open and
> let us elaborate a model of validity with a more formally defined
> semantics that allows to address specific instances of trains from a
> timetable.

That is somewhat covered by the below mentioned Trac ticket.

> Am 09.11.2012 10:27, schrieb Susanne Wunsch:
>> I filed a ticket for the issue of an "abstract operating period":
>>
>> http://trac.assembla.com/railML/ticket/187

Kind regards...
Susanne

--
Susanne Wunsch
Schema Coordinator: railML.common
Re: constraints for OperatingPeriod [message #917 is a reply to message #871] Wed, 06 February 2013 16:46 Go to previous message
Joachim Rubröder railML is currently offline  Joachim Rubröder railML
Messages: 0
Registered: November 2019
Dear all,
I totally agree about the different kinds of operatingPeriods and the need
for an 'abstract period'.

The second important point to consider is the operatingPeriod in the
different stages of the planning process. Evolving from an 'abstract
period' to a concrete 'bitmask'.

> Andreas Tanner <ata(at)ivude> writes:
>
> as far as railML 2.x is concerned, my suggestion was just to enhance
> the documentation.
>
> For railML 3.0, however, I would like to keep the discussion open and
> let us elaborate a model of validity with a more formally defined
> semantics that allows to address specific instances of trains from a
> timetable.

Many thanks for the discussion in this thread. I will be glad to implement
such a model of validity in a future version of railML.

I therefore switched the ticket to railML 3.0:
http://trac.assembla.com/railML/ticket/187

Kind regards,
Joachim

-------------------------------------
Joachim Rubröder
Schema Coordinator: railML.timetable


--
----== posted via PHP Headliner ==----
Previous Topic: Re: Steckenunterbruch/line blocking
Next Topic: Extension of places and service
Goto Forum:
  


Current Time: Fri Jul 12 19:21:11 CEST 2024