@edy What is your baseline browser? IE3, Netscape Navigator, MOSIAC, AoL, Compuserve... etc... or even the ancient text only Lynx (my first WWW experience). Or which HTML version are you targetting, before HTML 4?
Wow, I didn't realize how deep down the rabbit hole modern browsers have taken us. Reading up on it, I see that if you aren't keeping up with compatibility even a few years out or less, your site may break. I remember coding a webapp that used JQuery for some features and after an OS update on the phone, some of my app features didn't work. I then had to go back and figure out if it was JQuery, the update to the webkit engine, some function name/parameter changes that caused it. What a nightmare, and only so I could upload a "new" version of the same free app so that users wouldn't complain (not to mention having an old version still there for users who were still on the old OS version).

Anyways, back to the question...
Let's say I have an old Ubuntu 6 (dapper drake) Linux machine running some 11 year old version of Mozilla Firefox, or I have an old BlackBerry Curve phone I'm using to stream music (released on Rogers Wireless in Canada on Aug 4, 2010). Or even a BlackBerry Playbook tablet that was released 2011 and had the last OS update in 2012 which basically left it with a stale browser. Unfortunately I can't update these devices but they still operate perfectly fine and connect to the internet Wi-Fi no problem.
The goal is not to "scrape" sites. I understand most sites will not function properly. I just want a way to read a page and render it in a way that can be displayed on an old browser. Yes I can always use VNC on the device (like the Playbook) and just connect to a modern Linux machine with a VNC server and browser the internet that way. Perhaps the same for my old Ubuntu machine, if I can install an old version of VNC. However certain devices (especially "dead" ones without app support anymore) like phones would benefit from having an intermediary site that can render/reformat the page even in some way that lightens the bandwidth considerably and gets rid of much of the "bells and whistles" while still retaining at least the relevant information. That's the hard part.
For example, imagine I want to load up a news site. I could theoretically just grab the entire page, convert it to a static IMAGE and then just put interactive "hotspots" on the image where there are URL links, so that when hovered over it will allow you to click it and navigate. So the entire site isn't even text anymore... just graphical elements. But that wouldn't work for many other sites that do more complicated things. Obviously any kind of pop-up players wouldn't work, but I don't usually need them.
On the BlackBerry curve, I usually listen to music from internet radio stations. To get the IP address, I go to my favourite radio station on
www.streema.com, click on a station and then open up the PAGE SOURCE and find the URL of the stream. I can browse to that directly in my BlackBerry Curve browser it will play the station. However, the phone fails completely when it comes to even loading up the Streema.com website!
I'll give you an example... on streema.com I can search for Top 40 hits and find the following website pop-up with an embedded player:
http://streema.com/radios/play/American_Top_40_AT40Since none of this works on my BlackBerry Curve, I open source of that page and find the following URL in there:
<source src='
http://stream.revma.ihrhls.com/zc4802' type='audio/mpeg' />
I type that link into my BlackBerry Curve browser and it will give me the option to OPEN or SAVE. If I save it, it will start writing an mp3 file to local storage which will go on indefinitely (basically saving the music stream which I can play back later). Or I can choose OPEN and it will start playing it... I connect to line-out phono jack and have internet radio stream to some speakers. So while streema.com fails to load at all on my BlackBerry Curve (complains about unable to connect using current security settings), if it was formatted another way it would still be useful.
I think the problem will be that in order to pull out useful information, the reformatting for each site has to be customized and that will be an impossible task to make for a single "pass-through" site. That's probably the crux of the problem.