W3C home > Mailing lists > Public > whatwg@whatwg.org > December 2011

[whatwg] creating a new file via the File API

From: David Karger <karger@mit.edu>
Date: Mon, 19 Dec 2011 01:17:47 -0500
Message-ID: <4EEED70B.2020407@mit.edu>
When I run your example in chrome, all those links automatically 
download the file to the specified filename in my default download 
directory---none launch the file save dialog.  Of course that's because 
of how my chrome defaults are set.  And indeed I can right click and 
file-save-as.  But it's going to be a problem if I want to put a "save 
as" button on the page---having that result in download to a default 
directory because that's how chrome defaults are set isn't going to be 
the right behavior from the user's perspective.   Should there be a way 
to force open the save dialog, even if the default is to download to a 
fixed location?

On 12/19/2011 12:35 AM, Bronislav Klu?ka wrote:
> hi,
> if you look at the generated files examples, what you can see there 
> (again, only in chrome) is that
> 1/ I have some data in JS
> 2/ I create blobbuilder -> blob -> url to that blob
> 3/ I create a element with URL to that blob and  download attribute
> 4/ I initiate click on that link programmaticaly
>
> the result is is that Save file dialog is opened and when save/ok 
> button is hit, the blob data is stored in user selected file.
> Yes, I'm using download attribute, but URL is JS blob (local data).
> I do not see problem here. What are you missing?
>
> And yes, I also do not see security issues here, nothing user cannot 
> do today with regular download or programmer by uploading data to 
> server and then download them...
>
> B.
>
>
>
>
> On 19.12.2011 6:26, David Karger wrote:
>> Hi,
>>
>>    What you're doing is certainly connected, but I don't think it 
>> solves the problem I outlined.  Your approach allows specification of 
>> the download target as an attribute in html.   That's useful, but 
>> what's still missing, and I consider important, is a way to connect 
>> the html document to the "Save As" dialog available on all OSes. 
>> <input> tags lead browsers to launch the "Open File" dialog, which 
>> lets the user naturally navigate their file system to select a file 
>> to open.   Browsers also launch the analogous "Save As" dialog, but 
>> _only_ when you execute a download from a server.  I think it's 
>> important to enter the same "Save As" dialog programmatically, for 
>> client-side generated content.  I don't think this raises the 
>> security issues discussed at mozilla, because the user is engaged in 
>> the same interaction as they are on any other file download.
>>
>>
>> On 12/18/2011 11:13 PM, Bronislav Klu?ka wrote:
>>> Hi,
>>>
>>> This is quite crucial functionality and sadly not being addressed as 
>>> it would seem, because without it application cannot really be 
>>> applications (all you can do is to prepare data, upload those data 
>>> to server and let user download it manually by clicking somewhere, 
>>> which is annoying, unnecessary,  and quite frankly stupid) .
>>>
>>>
>>> but there is a way howto allow user to save file from javascript 
>>> without flash
>>>
>>> http://www.webnt.cz/demos/034_a_download/
>>>
>>> this demo (the generated files) allows you to download/drag'ndrop 
>>> generated file using JS (no flash)
>>> it's working in Chrome only at this point
>>> FF team is having some security issues I've been discussing with them
>>> https://bugzilla.mozilla.org/show_bug.cgi?id=676619
>>>
>>>
>>>
>>> B.
>>>
>>>
>>> On 16.12.2011 0:58, David Karger wrote:
>>>> It isn't clear to me that a "tag" question can be addressed by an 
>>>> "api" answer.   Even if there is an api for saving to file, isn't 
>>>> there value to being able to declare your intentions through a 
>>>> tag?  The <input type=file> tag specifies that a user will be able 
>>>> to interact to specify a file through a dialog.  There's absolutely 
>>>> no commitment that that file will actually be uploaded or input.  
>>>> That's up to the form or the javascript that handles the input.  It 
>>>> seems entirely consistent to be able to permit specification of a 
>>>> brand new file in that dialog that <input type="file"> is already 
>>>> creating.   What some javascript _does_ with the specified file 
>>>> might need to be implemented using a filesaver api, but that's 
>>>> separate from the declaration of an interaction for specifying the 
>>>> file.
>>>>
>>>> On 12/15/2011 6:45 PM, whatwg-request at lists.whatwg.org wrote:
>>>>
>>>> On Mon, 15 Aug 2011, David Karger wrote:
>>>>> /
>>>> />/  Apologies if I'm revisiting old territory.  I've been doing 
>>>> work on pure
>>>> />/  html/javascript applications that work entirely clientside
>>>> />/  (http://projects.csail.mit.edu/exhibit/Dido).  For 
>>>> persistence, they
>>>> />/  read and write local files.  There's already an<input 
>>>> type="file">
>>>> />/  interface for letting the user specify a file to be read.  And 
>>>> I can use
>>>> />/  the same interface, inappropriately, to let the user overwrite a
>>>> />/  preexisting file.  But things get much messier if I want to 
>>>> let the user
>>>> />/  specify a _new_ file to be written, because the file-open 
>>>> dialog doesn't
>>>> />/  offer users a way to specify a new filename.  What I'd like to 
>>>> be able
>>>> />/  to do is specify a tag, or a invoke some javascript method, 
>>>> that will
>>>> />/  produce the "save file" dialog typical of most systems, with a 
>>>> graphical
>>>> />/  directory browser but including the option to specify a new 
>>>> filename.
>>>> />/  This problem isn't unique to me; a discussion on stackoverflow 
>>>> appears
>>>> />/  at
>>>> />/  
>>>> http://stackoverflow.com/questions/2897619/using-html5-javascript-to-generate-and-save-a-file
>>>> />/  where the proposed solution is to use flash---and that would 
>>>> be an
>>>> />/  unfortunate loss of html5 purity.  They also suggest the hack 
>>>> of using a
>>>> />/  data: url but that has size limitations.
>>>> />/
>>>> />/  Perhaps<input type="file">  could be given an attribute 
>>>> specifying
>>>> />/  whether a new filename is permitted?
>>>> /
>>>> On Wed, 7 Sep 2011, Eric U wrote:
>>>>> /
>>>> />/  This sounds like a job for the FileSaver interface.  Currently no
>>>> />/  browser implements it, but we at Chrome have been considering 
>>>> it.  At
>>>> />/  TPAC last year we discussed it a bit in the WebApps WG 
>>>> meeting; IIRC we
>>>> />/  talked about letting it take a URL instead of or in addition 
>>>> to just a
>>>> />/  Blob, for more general utility.
>>>> />/
>>>> />/  I suggest you bring it up on public-webapps@, where that spec 
>>>> lives.
>>>> />/  
>>>> http://dev.w3.org/2009/dap/file-system/file-writer.html#idl-def-FileSaver
>>>> /
>>>> I agree that an API like FileSaver is the right way to do this. Using
>>>> <input type=file>  wouldn't really fit well because that's more for
>>>> providing data for upload than providing a file for writing.
>>>>
>>>
>
Received on Sunday, 18 December 2011 22:17:47 UTC

This archive was generated by hypermail 2.4.0 : Wednesday, 22 January 2020 16:59:38 UTC