Re: [Edbrowse-dev] Images

[Adam Thompson <arthompson1990@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2724 bytes --]

On Sun, Apr 27, 2014 at 03:37:22PM +0000, Karl Dahlke wrote:
> > Webvism is both a service and a Firefox plugin for solving captchas.
> 
> I'm guessing it sends the img url to some website, maybe with yourcredentials,
> perhaps as cookies, then the site posts it, some user solves it,
> and returns the letters to you, which you type in,
> or cut&paste in, or maybe fills in automatically.
> 

Either that or the image itself.  If it's used in realtime then you probably need to send the downloaded image as lots of captures generate different images each time they're loaded.

> An edbrowse function could do all of this,
> if we knew how it worked inside, except, there is no easy way to get the url.
> 
> As of today, the alt text of an image is in brackets,
> unless of course that image is a clickable url whence it is in braces.
> I'm not thrilled about brackets since text is sometimes in brackets anyways.
> Rarely though is text naturally in braces or angles <>.
> So brackets are ok I guess for an image but if you have other suggestions I'm open.

I'm fairly used to brackets, I think this comes from when I used to browse the
web with the links2 browser though so I don't mind. As long as any change is documented.
> Maybe ~picture of a house~.
> Those are eused less frequently.

That also works.

> 
> Anyways let's say this is in place, then you want the edbrowse function
> to pull out the url from the src= attribute, but oops,
> this is entirely lost through browsing.
> It's not clickable, so I didn't particularly care about it.
> Visual browsers however let you right click on a picture
> to download it or whatever.
> So imagine that I carried some internal tags for the image, like I do with hyperlinks
> and form fields etc.
> If these tags were present then the A command could bring up the url for that picture,
> just like it does for links and so on.
> Then the edbrowse function could grab the url and send it to Webvism
> and scrape the letters off and ^ back and maybe even put them into the next input field.
> Feasible, if I retained the tags and src reference for images.
> Wouldn't be a big job, but not trivial and probably not our highest priority.

It would certainly have its uses, but as you say it's not really high priority at the moment.

As for the webvism function, if it does indeed use the url rather than the
image then I wonder if one could unbrowse the page, search for the alt text,
use a regex to extract out the url and send it to the site that way?

This approach wouldn't be very robust or nice to implement but could be made to
work in cases where some details about the image's alt text are known.

Cheers,
Adam.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 473 bytes --]