Re: Multi-GET, extreme compression?

["Nicolas Mailhot" <[e-mail]>]

Le Mer 27 février 2013 17:28, Julian Reschke a écrit :
> On 2013-02-27 17:20, Julian Reschke wrote:
>> On 2013-02-27 16:40, Nicolas Mailhot wrote:
>>> Julian Reschke <julian.reschke@...> writes:
>>>
>>>>
>>>> On 2013-02-27 11:16, Nicolas Mailhot wrote:
>>>>> James M Snell <jasnell@...> writes:
>>>>>
>>>>>
>>>>>> Fair enough
>>>>>
>>>>> There is a similar need for mput/mpost, the fact current web apps
>>>>> require
>>>>> a separate user sequence for each file a user wants to
>>>>> publish/attach to a
>>>>> message is one of the few remaining use-cases where they suck
>>>>> compared to
>>>>> local apps.
>>>>
>>>> But that's UI (HTML/JS), not protocol, right? Also, that's solvable; I
>>>> happen to be in a project where our web app supports bulk upload of
>>>> files by drag & drop to the browser window...
>>>
>>> That's just another workaround, where you paper over the missing
>>> feature with
>>> gobs of site-specific javascript and by pretending the average user
>>> drag-and-drops files in apps. I've seen the same demoware years ago it
>>> does not
>>> work out in real life.
>>>
>>> The average user does not drag and drop he uses the file selector,
>>> that maps to
>>> standard html forms, that maps to what the protocol knows to do (one
>>> element at
>>> a time). In browsers the file selector is restricted to single file
>>> selection to
>>> respect what non-js-extended http/html ecosystem allows.
>>>
>>> And, lastly, most web app developpers will not bother with loads of
>>> feel-good js
>>> workarounds since what the users want is their default file selector,
>>> not some
>>> kind of js emulation.
>>
>> I still don't understand what the constraints of the browsers have to do
>> with the protocol. Browsers use multipart uploads, and those allow
>> multiple files already. What am I missing?
>>
>> Best regards, Julian
>
> Also, see
> <http://www.w3.org/TR/html-markup/input.file.html#input.file.attrs.multiple>.
> No JS needed at all.

Ah, nice, someone finally noticed this part was broken. Pity it's not
deployed yet.
Though the html5 element would still map naturally to an http/2 mput :)

Thanks for the pointer!

-- 
Nicolas Mailhot