RE: [FileReader API, ProgressEvents] Design patterns, FileWriter API

2009-11-19 Thread Marcin Hanclik
Hi Arun,

 To be clear, IMHO it's absolutely too late for FileReader
FileReader is still ED, therefore we may have time, I think.

Regarding FileWriter, I'm open to considering new event names, but in
general, discussing FileWriter's event model may be putting the cart in
front of the horse.  Even if we had an event-driven FileWriter, what
would it do?  We'll need to figure out filesystem access (real or
virtual) first! :-)
Agreed.
I have been thinking in terms of API design patterns and IMHO their target is 
to first document what we have in order to be able to foresee potential 
incompatibilities.
That is why I shout early :)

Thanks,
Marcin

Marcin Hanclik
ACCESS Systems Germany GmbH
Tel: +49-208-8290-6452  |  Fax: +49-208-8290-6465
Mobile: +49-163-8290-646
E-Mail: marcin.hanc...@access-company.com

-Original Message-
From: Arun Ranganathan [mailto:a...@mozilla.com]
Sent: Thursday, November 19, 2009 12:55 AM
To: Marcin Hanclik
Cc: Anne van Kesteren; WebApps WG; public-device-a...@w3.org
Subject: Re: [FileReader API, ProgressEvents] Design patterns, FileWriter API

Marcin Hanclik wrote:
 Hi Anne,

 XHR still is used also for data retrieval, so it is a kind of border case and 
 I can live with load there :) .

 Using load for writing to a file will mean that we are stuck with the 
 legacy stuff. load and write pull semantically in the opposite 
 directions, IMHO.

 I think there is still time to change it in case of FileReader and prepare 
 background for FileWriter.


To be clear, IMHO it's absolutely too late for FileReader, and also not
desirable.  I've already mentioned developer familiarity with onload,
etc. and I strongly disagree with changing names in this case.  The
existing progress events are fine for FileReader.  We've also got a beta
implementation in place in Fx 3.6.

Regarding FileWriter, I'm open to considering new event names, but in
general, discussing FileWriter's event model may be putting the cart in
front of the horse.  Even if we had an event-driven FileWriter, what
would it do?  We'll need to figure out filesystem access (real or
virtual) first! :-)  More on this topic in separate email.

-- A*



Access Systems Germany GmbH
Essener Strasse 5  |  D-46047 Oberhausen
HRB 13548 Amtsgericht Duisburg
Geschaeftsfuehrer: Michel Piquemal, Tomonori Watanabe, Yusuke Kanda

www.access-company.com

CONFIDENTIALITY NOTICE
This e-mail and any attachments hereto may contain information that is 
privileged or confidential, and is intended for use only by the
individual or entity to which it is addressed. Any disclosure, copying or 
distribution of the information by anyone else is strictly prohibited.
If you have received this document in error, please notify us promptly by 
responding to this e-mail. Thank you.


Re: [FileReader API, ProgressEvents] Design patterns, FileWriter API

2009-11-19 Thread Robin Berjon
On Nov 19, 2009, at 11:35 , Marcin Hanclik wrote:
 To be clear, IMHO it's absolutely too late for FileReader
 FileReader is still ED, therefore we may have time, I think.

Actually, it's published (as a WD) and has a rather long background history of 
previous development in the File Upload spec. I think I'm not the only one who 
would like this particular baby to move forward rather fast.

-- 
Robin Berjon - http://berjon.com/






Re: [FileReader API, ProgressEvents] Design patterns, FileWriter API

2009-11-18 Thread Anne van Kesteren
On Wed, 18 Nov 2009 02:30:16 +0100, Arun Ranganathan a...@mozilla.com  
wrote:
I think that just as the names 'load*' were chosen for generic data  
transfer events (either networked or within a document), and are used  
within documents loaded in the DOM, XHR, and FileReader, we'll need  
reusable 'write*' events.  Without bikeshedding too much, I like your  
proposal above, but wonder whether we should use the name 'write*' or  
something else.  Since we already have document.write, 'write' is  
probably the most vetted string to use here :)


For what it's worth, for XMLHttpRequest sending events (which are  
arguably somewhat like write) we still use loadstart/... and simply  
dispatch them on a distinct object. I've no idea what the file writer API  
will look like, but I can imagine we might be able to do the same there.



--
Anne van Kesteren
http://annevankesteren.nl/



RE: [FileReader API, ProgressEvents] Design patterns, FileWriter API

2009-11-18 Thread Marcin Hanclik
Hi Anne,

XHR still is used also for data retrieval, so it is a kind of border case and I 
can live with load there :) .

Using load for writing to a file will mean that we are stuck with the legacy 
stuff. load and write pull semantically in the opposite directions, IMHO.

I think there is still time to change it in case of FileReader and prepare 
background for FileWriter.
I like the ProgressEvents spec and would keep it generic, i.e. change the names 
there to be generic and mandate their change in each spec that refer to 
ProgressEvents. PE is perfect to be a design pattern for asynchronous (and 
lengthy, more-than-one-shot, with-start-and-end, abortable and potentially 
erroneous) DAP APIs.

In the firing grammar:
progressgrammar = loadstart working (error | abort | load) loadend
working = *( progress | (progress suspend progress) | (progress stalled 
progress) )

potentially rewritten to

progressgrammar = start working (error | abort | done) end
working = *( progress | (progress suspend progress) | (progress stalled 
progress) )

we would only need to change the working rule to accommodate various event 
firing scenarios. Thus under the same design pattern we could accommodate XHR, 
HTML5's media API, FileReader and any new DAP API.
The event names could be related to the API for naming consistency, but firing 
model would be pre-defined were possible for design consistency.

For example for directory walker (aka File Directory API or File API Level 3: 
Directories as proposed by Robin) we could have:

directorywalkereventgrammar = start working (error | abort | done) end
working = *( enterdir *file leavedir )

to be able to walk over e.g. 1M file entries in some FS.

Any thoughts?
I guess it may be over-engineering, but ...

Thanks,
Marcin

Marcin Hanclik
ACCESS Systems Germany GmbH
Tel: +49-208-8290-6452  |  Fax: +49-208-8290-6465
Mobile: +49-163-8290-646
E-Mail: marcin.hanc...@access-company.com

-Original Message-
From: Anne van Kesteren [mailto:ann...@opera.com]
Sent: Wednesday, November 18, 2009 10:31 AM
To: a...@mozilla.com; Marcin Hanclik
Cc: WebApps WG; public-device-a...@w3.org
Subject: Re: [FileReader API, ProgressEvents] Design patterns, FileWriter API

On Wed, 18 Nov 2009 02:30:16 +0100, Arun Ranganathan a...@mozilla.com
wrote:
 I think that just as the names 'load*' were chosen for generic data
 transfer events (either networked or within a document), and are used
 within documents loaded in the DOM, XHR, and FileReader, we'll need
 reusable 'write*' events.  Without bikeshedding too much, I like your
 proposal above, but wonder whether we should use the name 'write*' or
 something else.  Since we already have document.write, 'write' is
 probably the most vetted string to use here :)

For what it's worth, for XMLHttpRequest sending events (which are
arguably somewhat like write) we still use loadstart/... and simply
dispatch them on a distinct object. I've no idea what the file writer API
will look like, but I can imagine we might be able to do the same there.


--
Anne van Kesteren
http://annevankesteren.nl/



Access Systems Germany GmbH
Essener Strasse 5  |  D-46047 Oberhausen
HRB 13548 Amtsgericht Duisburg
Geschaeftsfuehrer: Michel Piquemal, Tomonori Watanabe, Yusuke Kanda

www.access-company.com

CONFIDENTIALITY NOTICE
This e-mail and any attachments hereto may contain information that is 
privileged or confidential, and is intended for use only by the
individual or entity to which it is addressed. Any disclosure, copying or 
distribution of the information by anyone else is strictly prohibited.
If you have received this document in error, please notify us promptly by 
responding to this e-mail. Thank you.


Re: [FileReader API, ProgressEvents] Design patterns, FileWriter API

2009-11-18 Thread Arun Ranganathan

Marcin Hanclik wrote:

Hi Anne,

XHR still is used also for data retrieval, so it is a kind of border case and I can live 
with load there :) .

Using load for writing to a file will mean that we are stuck with the legacy stuff. 
load and write pull semantically in the opposite directions, IMHO.

I think there is still time to change it in case of FileReader and prepare 
background for FileWriter.
  


To be clear, IMHO it's absolutely too late for FileReader, and also not 
desirable.  I've already mentioned developer familiarity with onload, 
etc. and I strongly disagree with changing names in this case.  The 
existing progress events are fine for FileReader.  We've also got a beta 
implementation in place in Fx 3.6.


Regarding FileWriter, I'm open to considering new event names, but in 
general, discussing FileWriter's event model may be putting the cart in 
front of the horse.  Even if we had an event-driven FileWriter, what 
would it do?  We'll need to figure out filesystem access (real or 
virtual) first! :-)  More on this topic in separate email.


-- A*



Re: [FileReader API, ProgressEvents] Design patterns, FileWriter API

2009-11-18 Thread Charles McCathieNevile
On Wed, 18 Nov 2009 02:30:16 +0100, Arun Ranganathan a...@mozilla.com  
wrote:



Greetings Marcin,

Thanks for the thoughtful feedback.  My comments below:

In my opinion some part of the design from ProgressEvents is taken over  
in FileReader API too directly.

Specifically the event names are the same as within the ProgressEvents,

...


What's in a name? [Floral poetry reference elided :-) ].

You are right to point out that there's some inconsistency in the naming  
convention.  This discussion came up with the discussion of readyState  
[1] state name changes, when READING was considered as a potential  
readyState, were it not for consistency with the progress events in  
question (loadstart, load, loadend... ).  We elected to stick to LOADING  
to match the names of existing progress events with 'load' in them.

...



To date, you'll note that progress event nomenclature reflects loading  
or reading operations, since there are very few write metaphors on  
the web that have affiliated events bound to them.


Actually, this is a false etiology. The naming conventions are based on  
what people were already implementing (as Robin says, it is often easier  
to extend the definition than change the term, as linguistics shows us  
over a far longer history than the Web).


 If the design of the FileWriter is similar to the design of the  
FileReader (which is something we're currently working on), then I think  
your names make sense.


I don't actually. We have dumb names, because those are the terms we keep  
using... and the ones people are used to.


Then, the ProgressEvents spec could act as a design pattern definition  
for lengthy, asynchronous operations.
To make it happen, the names of the events there could be changed to be  
generic:




I think that just as the names 'load*' were chosen for generic data  
transfer events (either networked or within a document), and are used  
within documents loaded in the DOM, XHR, and FileReader, we'll need  
reusable 'write*' events


Actually, if the argument to keep the names makes sense, then it makes  
sense, so there is not much point in trying to separate out a bunch of  
things with a new name. Experience has shown it doesn't usually work  
anyway.



loadstart - start
progress
stalled
suspend
error


I like the proposal to have a section specific to data transfer and  
loading, but am wary of splitting specs.  Input and feedback from the  
author of the ProgressEvents specification would be welcome here.


The author is the Working Group - and ergo all the members. I'm just the  
editor.


cheers

Chaals

--
Charles McCathieNevile  Opera Software, Standards Group
je parle français -- hablo español -- jeg lærer norsk
http://my.opera.com/chaals   Try Opera: http://www.opera.com



Re: [FileReader API, ProgressEvents] Design patterns, FileWriter API

2009-11-17 Thread Arun Ranganathan

Greetings Marcin,

Thanks for the thoughtful feedback.  My comments below:


In my opinion some part of the design from ProgressEvents is taken over in 
FileReader API too directly.
Specifically the event names are the same as within the ProgressEvents, but I 
assume they should be adjusted to the FileReader API.
Therefore instead of (forgetting the above issue with the callback model for 
now):

  attribute Function onloadstart;
  attribute Function onprogress;
  attribute Function onload;
  attribute Function onabort;
  attribute Function onerror;
  attribute Function onloadend;

we could have:

  attribute Function onreadstart;
  attribute Function onprogress;
  attribute Function onread;
  attribute Function onabort;
  attribute Function onerror;
  attribute Function onreadend;
  



What's in a name? [Floral poetry reference elided :-) ].

You are right to point out that there's some inconsistency in the naming 
convention.  This discussion came up with the discussion of readyState 
[1] state name changes, when READING was considered as a potential 
readyState, were it not for consistency with the progress events in 
question (loadstart, load, loadend... ).  We elected to stick to LOADING 
to match the names of existing progress events with 'load' in them.


I'll note that 'onload' is very ubiquitous on the web platform, and has 
been for a while now.  These progress events are also used within XHR.  
This kind of consistency is very desirable for developer convenience, 
and in my opinion, trumps the perfect name consideration.  We've used 
consistency in naming before in this specification.  Here are a few 
other examples:


1. INTIAL -- EMPTY, for closer name similarity to HTMLMediaElement's 
network status.

2. mediaType -- type, for closer name similarity with style.type
3. Error conditions like NOT_FOUND_ERR, SECURITY_ERR, and ABORT_ERR are 
used in DOMException codes and XHR's error codes respectively, including 
with the same error code.


Essentially, the argument here is to reuse what developers are familiar 
with.  *In particular* I'm swayed by the pervasive use of 'onload' on 
the web.  However, I like your next point:



Assuming that we will have an interface like FileWriter in the (near) future, 
we could already now anticipate that the interface would include e.g. the 
following:

  attribute Function onwritestart;
  attribute Function onprogress;
  attribute Function onwrite;
  attribute Function onabort;
  attribute Function onerror;
  attribute Function onwriteend;
  



To date, you'll note that progress event nomenclature reflects loading 
or reading operations, since there are very few write metaphors on 
the web that have affiliated events bound to them.  If the design of the 
FileWriter is similar to the design of the FileReader (which is 
something we're currently working on), then I think your names make sense.



Then, the ProgressEvents spec could act as a design pattern definition for 
lengthy, asynchronous operations.
To make it happen, the names of the events there could be changed to be generic:
  


I think that just as the names 'load*' were chosen for generic data 
transfer events (either networked or within a document), and are used 
within documents loaded in the DOM, XHR, and FileReader, we'll need 
reusable 'write*' events.  Without bikeshedding too much, I like your 
proposal above, but wonder whether we should use the name 'write*' or 
something else.  Since we already have document.write, 'write' is 
probably the most vetted string to use here :)



loadstart - start
progress
stalled
suspend
error
abort
load - done
loadend - end
  



Sure.  On this question I'm less opinionated.

Additionally the ProgressEvents spec could be divided (or split into two 
documents? ) to contain the section specific to the design pattern definition 
and to the section specific to data transfer / loading.
  


I like the proposal to have a section specific to data transfer and 
loading, but am wary of splitting specs.  Input and feedback from the 
author of the ProgressEvents specification would be welcome here.


-- A*

[1] http://lists.w3.org/Archives/Public/public-webapps/2009OctDec/0691.html