- From: Cecile Roisin <Cecile.Roisin@imag.fr>
- Date: Tue, 23 Dec 1997 12:09:33 +0100
- To: hoschka@w3.org
- cc: Cecile.Roisin@imag.fr, Muriel.Jourdan@imag.fr, www-multimedia@w3.org
[SORRY, THIS MAIL REPLACES THE PRECEDENT ONE] Dear SMIL editor, We carefully read the SMIL Working Draft and we have the following remarks on that document. These remarks mostly aim at clarifying the concepts and the sematics of the proposal. Indeed, this proposal still contains numerous ambiguities which can have a lot of drawbacks in its future use (different interpretations of the same document). In a next paper, we will discuss about the authoring capabilities of SMIL and for that purpose, we will use the example given at the end of the SMIL Working Draft. In the following list, the remaks are ordered by the linear reading of the draft and we have tried to systematically refer to the sections numbers. Remark 1 Section 6.1 General semantics One case of end time of an element E is missing: when E is a composite element, its end can be determined by the end of its descendants (example: a PAR element without attribute will end when the child of longest duration ends). Remark 2 Section 6.2 General Semantics End Time No semantics is given for the situation when one (or more but not all) of the children has an unknown end time. As far as we understand, the definition should be given as: "By default, the end time of a parallel element is equal to the maximum end-time of all children having a known duration in the parallel element". Moreover, at this stage of the draft, the notion of known/unknown end time is not clear. One solution to improve this is to put section 6.4 (Media objects elements) before section 6.2 and to explain when the duration of an object is unknown. It should also be said how this property is propagated through composition: in which cases a composed object built with a seq or a par has a unknown duration? From our understanding, there are two cases of basic objects with unknown duration: either static objects (text, image) with no duration attribute (dur) or objects with the repetition attribute (repeat) set to 0. There is a lot of ways to build parallel objects with unknown duration, such as the three following ones. 1) <par endsync = "last" > <o1 .... /> <o2 ..../> </par> with o1 and o2 two unknown duration objects 2) <par endsync = "first" > <o1 .... /> <o2 ..../> </par> o1 and o2 have both unknown duration 3) <par endsync = o1 > <o1 .... /> <o2 ..../> </par> with o1 a unknown duration object Always from our understanding, a composite object built with a sequence operator has an unknown duration if and only if one of its children has an unknown duration. Remark 3 Section 6.2 Attributes In order to be more precise, we propose: endsync (optional) "last" : The parallel element ends at the same time as the child with the maximum end time of all children having a known duration in the parallel group. This is the default semantics of a parallel schedule (i. e. when a parallel element has no endsync attribute). "first" : The parallel element ends at the same time as the child with the minimum end time of all children having a known duration in the sequential group. Remark 4 Section 6.2 Attributes The reference " (see 6.6) " is wrong. If a parallel element contains both an "endsync" attribute and an "end" attribute (see 6.5), the element ends at the minimum of the end times specified by these two attributes. Remark 5 Section 6.2 Attributes dur (optional) What is the semantics of a parallel element with such a duration attribute? (see remark 12 for general comments about the problem of ambiguities when there are several attributes to define end or begining points). Remark 6 Section 6.2 Attributes repeat (optional) What is the semantics if the element has both dur and repeat attributes? (see Remark 12) Remark 7 Section 6.4 It should be placed before section 6.2 (see Remark 2) Remark 8 Section 6.4 General Semantics What is exactly the semantics of ref media object? Can it be used for other object types than the list "audio", "video", "text" and "img", for example a Delay type? Remark 9 Section 6.4 Attributes dur (optional) Explanations about potential conflicts between duration attributes and inherent duration of continous objects are given too late from this point of the draft. Remark 10 Section 6.5 General Semantics This part should only contain the two first sentences of the first paragraph: "The synchronization attributes "begin" and "end" can be added to any schedule element. These attributes change the default begin and end times of the element." Then, we propose to place the Attributes part (definition of Begin and End attributes) which is after the "Error Handling" paragraph in the document. The rest of this section will be more clear since its aim is to specify these two attributes. Remark 11 Section 6.5 General Semantics What is the semantics of an end synchronization attribute specified as an offset value? There are only explanations and examples given for the other case: an offset value applied to the begin point of an element. Remark 12 Section 6.5 Error Handling This section should be more complete and explicit in the description of invalid documents. Some parts should be outside the section 6.5 because it is also related with other sections (see remarks 5 and 6). Something general about this document, is that there are not enough explanations on potential conflicts between several ways to define end/begin time of an element. Some cases are stated: section 6.2 :"The default end time of a parallel element can be overriden by using the endsync, the dur or the end attribute" Error Handling : (Rule 1) "If an element contains begin, end and dur attributes the dur attributes is ignored" (Rule 2) "If a continuous media object has either a dur attribute or a begin and end attribute, the elements duration is the minimum of its inherent duration and the duration defined by the attributes" (Rule 3) " ... possible errors are ... conficts between begin/end time and the duration of an object. Is the first rule an exception of the third one ? Suppose I have an continous object with an inherent duration of 30s, if I write <o1 begin=0 end =30 dur= 40> or < o1 dur = 40> I do not have the same semantics. A proposition which seems to us more simple to manage such problems: only one rule: in case of conflicts, the duration of an element is always the minimum one. Let's take some examples (The question is: is this example valid or not, if valid what is its ending time ?): 1) <par endsync = "first" dur = 20> <o1 .... /> <o2 .... /> </par> 2) <par endsync = "first" end = id(objectZ) (begin)> <o1 .... /> <o2 .... /> </par> objectZ is a sibling object of the composite one. 3) - lipsync attribute together with dur attribute 4) <par repeat 4, dur = 30 > <o1 dur 20 /> <o2 /> </par> o1 and o2 have both unknown durations Remark 14 Section 7.1. SMIL Basic Layout This approach for specifying the layout is too basic for taking into account needs for moving elements, spatio-temporal specifications etc. Some remarks about the example given at the end of the document: Remarks 15: In the statement : "The text Web Growth is shown once the image has reached its final position". What about the notion of final position? In the smil document: Remarks 16: duration of graph is 45s and not 60s as shown in the figure Remarks 17: the anchor is applied only to tim-video and not as specified initially to both the text and the video Remarks 18 : the figure suggests that the end of the audio and video objects are synchronized but this is not what it is done in the smil specification Muriel and Cecile ---------------------------------------------------------- Muriel Jourdan Cecile Roisin EMAIL: Muriel.Jourdan@inrialpes.fr, Cecile.Roisin@inrialpes.fr --------------------------- Unite de Recherche INRIA Rhone-Alpes - projet OPERA ( http://opera.inrialpes.fr/OPERA/ ) 655 avenue de l'Europe F-38330 Montbonnot Saint-Martin TEL: (33) 04 76 61 53 60 FAX: (33) 04 76 61 52 07 -----------------------------------------------------------
Received on Tuesday, 23 December 1997 06:11:03 UTC