- From: Alfonso Martínez de Lizarrondo <amla70@gmail.com>
- Date: Thu, 23 Jul 2009 00:04:13 +0200
- To: public-webapps@w3.org
Although the thread is almost a year old, I haven't found any replies to these questions, so I'm sharing what I've found: > I have a few questions and an idea. > > Does calling send( file ) generate the appropriate Content-Type > request header with matching boundary? No, it doesn't. I've filed https://bugs.webkit.org/show_bug.cgi?id=26979 about it. > How can this feature be detected? I don't think that it can be detected. You must know that the visitor is using the correct webkit version. > Idea: > How about passing a form to send? When passing a form to send, the > correct headers are set from the form's enctype, generating the > boundary, content-disposition and other stuff for multipart requests. Yes, that's what I thought that day after sleeping. That would be great, much better than the option of Firefox 3.5 to read the whole file as a string and send that. The browser takes care of reading the file and sending it as it is needed, like any current form without extra memory requirements. > The only problem is that it complicates send with much functionality. > Responsibly adaptive and well designed web apps will want to creat > feature tests and fallback strategies, but this will be awkward and > difficult to do, given the current API. Why not create a separate > method - sendForm. > > if(xhr.sendForm) > xhr.sendForm(form); > > Garrett That's exactly the proposal that I wanted to present to the Web Apps, if it's possible to reuse exactly the same server code, but provide an UI at the browser then lots of people will gladly use it. Just think that basically the code to degrade gracefully would be if (xhr.sendForm) xhr.sendForm(form); else form.send(); Regards
Received on Wednesday, 22 July 2009 22:04:53 UTC