Subject: Re: SDL-News: A question about history in SDL-2000
From: Eckhardt Holz (holz#informatik.hu-berlin.de)
Date: Thu Dec 20 2001 - 15:59:34 GMT
Dear all,
after checking my memory and some files on my computer I`ll give some
short answers here:
On 12/19/2001 4:36 PM, Rick Reed TSE wrote:
> Dear Birger,
>
> In Geneva in September you volunteered to provide an answer to the
> following:
>
>> From: Jacqueline Floch <Jacqueline.Floch#item.ntnu.no>
>> Reply-To: Jacqueline Floch <Jacqueline.Floch#item.ntnu.no>
>> Date: Tue, 06 Nov 2001 16:06:49 +0100
(...)
>>
>>
>> First I lack the recommendation to describe what happens when a composite
>> state that is not a state aggregation is re-entered through a nextstate
>> with history. Does the description given for state aggregation also apply
>> for any composite state?
>>
Yes, a nextstate with a history applies to any composite state. It has a
deep history semantics, i.e. the innermost state(s) where the transition
did originate will be the target of the nextstate.
Dash-nextstate on the other side is a shallow history, i.e. the target
is the (composite) state where the input-part of transition is directly
attached to.
>> Assuming this, I would like to use history for describing the suspension
>> and resumption of a composite state (and its substates). Suspension and
>> resumption may be easliy described using "suspend", "resume" messages. When
>> receiving a suspend message, a composite state would transit to a suspended
>> state waiting for a resume message that would bring the state machine back
>> to the composite state with history - i.e. with all substates in the same
>> state as before suspension. It is not possible to describe suspension in
>> that way using SDL 2000 because of the limitation set in 11.12.2.1 (see
>> above): if history is used, the next state must be the state in which the
>> transition is activated. I wonder why this limitation has been set by the
>> study group 10? I lack support for re-entering a composite state with
>> history at any time in a state graph.
We have discussed this issue at least by e-mail (cf. mails concerning
TDA13 Item 18, Oslo(?)-Meeting), but decided to be consistent with the
existing dash-nextstate. It is therefore not possible to suspend, go to
some other state and later resume.
Anyway - I will think about the suspend-resume-issue it little bit more,
may be there is some other solution.
>>
>> When re-entering a state, I found it difficult to deal with the entry
>> procedure. I intended to describe initialisation actions in the entry
>> procedure. Re-entering a state does not necessarily require the same
>> initial actions to be performed as when entering a state. I lack a
>> "re-entry" procedure or at least some basic support in SDL for testing the
>> type of entry in a state (i.e. enter vs. re-enter). I wonder if this has
>> been discussed in the study group 10?
Here SDL follows the same path as UML, if you enter some internal state,
you have to perform all the exits and entries along your way. History
remembers the last state of the statemachine (where you left it), not
the state of your local data (variables etc.)
>>
>> Thank you in advance for feedback,
>> Jacqueline Floch.
>
> When can we expect to see this answer?
>
> --
> Rick Reed - rickreed#tseng.co.uk
> Tel:+44 1455 55 96 55 Fax:+44 1455 55 96 58 Mob.:+44 7970 50 96 50
>
I wish you all a merry christmas & happy new year
Eckhardt Holz
>
This archive was generated by hypermail 2a23 : Mon May 05 2008 - 20:30:55 GMT