Isn't this how the internet was supposed to work in the first place? I remember Netscape navigator having a 'go offline' icon in the corner.
In Netscape days, many people would have to pay by the minute to be connected to the internet. In those days, web pages generally contained far more information than they do now, and were less interactive. So you'd connect, load the content you wanted to see, disconnect, and then just sit there and read it for free, instead of bleeding cash.
These browsers were born in the era of dialup Internet that had per minute charges and/or long distance charges. At the very least you were tying up your family's phone line.
Basically it's like paying for every minute your cable modem is plugged in.
For the feature itself: Netscape had integration with the modem connectivity for the OS and would initiate a connection when you tried to visit a remote page. Offline mode let you disable automatic dialing of the modem.
> Netscape had integration with the modem connectivity for the OS and would initiate a connection when you tried to visit a remote page.
That's not "integration with modem connectivity", that's just going through the OS's socket API (or userland socket stack, e.g. Trumpet Winsock); where the socket library dials the modem to serve the first bind(2). Sort of like auto-mounting a network share to serve a VFS open(2).
Try it yourself: boot up a Windows 95 OSR2 machine with a (configured) modem, and try e.g. loading your Outlook Express email. The modem will dial. It's a feature of the socket stack.
These socket stacks would also automatically hang up the modem if the stack was idle (= no open sockets) for long enough.
My point was that a quiescent HTML4 browser has no open sockets, whether or not it's intentionally "offline." If you do as you say — load up a bunch of pages, and then sit there reading them — your modem will hang up, whether or not you play with Netscape's toggles.
(On single-tasking OSes like DOS — where a TCP/IP socket stack would be a part of a program, rather than a part of the OS — there was software that would eagerly hang up the modem whenever its internal socket refcount dropped to zero. But this isn't really a useful strategy for a multitasking OS, since a lot of things — e.g. AOL's chatroom software presaging AIM — would love to poll just often enough to cause the line that had just disconnected to reconnect. Since calls were charged per-minute rather than per-second, these reconnects had overhead costs!)
> [Netscape's] offline mode let you disable automatic dialing of the modem.
When you do... what?
When you first open the browser, to avoid loading your home page? (I guess that's sensible, especially if you're using Netscape in its capacity as an email client to read your already-synced email; or using it to author and test HTML; or using it to read local HTML documentation. And yet, not too sensible, since you need to open the browser to get to that toggle... is this a thing you had to think about in advance, like turning off your AC before shutting off your car?)
But I think you're implying that it's for when you try to navigate to a URL in the address bar, or click a link.
In which case, would the page, in fact, be served from the client-side cache, or would you just get nothing? (Was HTTP client-side caching even a thing in the early 90s? Did disks have the room to hold client-side caches? Did web servers by-and-large bother to send HTTP/1.0 Expires and Last-Modified headers? Etc.)
IIRC, some of the browsers, in "offline mode", would actually serve pages from the local cache, not even attempting to fetch them from the remote server. If you attempted to navigate to a page that hadn't been cached, you got an error of some kind.
I was considered mad at the time for upping my cache to a whopping 2Mb by friends, but Netscape's cache was highly configurable.
Things like cache pages, but not images, always cache bookmarked pages, cache iframe pages (which were often navigation), etc. Netscape 4 added CSS to that mix.
Browser shows “I’m offline. Do you want to proceed?” dialog. If OK, tries to open socket. Upon such request, OS will try to dial up on modem with PPP, which takes ~30s. PPP session disconnects after few minutes. Else abort. If browser is online in the first place, skip that dialog part.
- Was caching a thing? Did disks have room to hold cache?
Oh it was, there were rooms. Alas, 56kbps(7kB/s) ideal means each pages are kilobytes to single megabyte large at most. ADSL connections at ~1Mbps(1/8 MB/s) means a 256kB page loads in 1 second on a fine day. Even a 200MB disk space holds 1024 of such ~200kB pages, and IE back in those days had a hidden attributes but ordinary folder deep down from C: that holds about that large random cache of files that it decided to hold.
Fuck I’m old now
[0] https://dev.opera.com/blog/event-streaming-to-web-browsers/
I would have said UUCP for news but that's showing my age.
The browser cache used to actually be quite dependable as an offline way to view pages but this seems to have fallen out of favour in the mid naughties. I remember how disgusted I was when I realised Safari was no longer me letting see a page unless it could contact the server and download the latest version.
I used to have a caching proxy server that would basically MITM my browsing and be more vigilant than even the cache and it really worked quite well. This was back in the 90s when every bit of your max 54kbs counted, or when you wanted to read something while your Dad or sister wanted to also use the phone.
Anyway, you can no longer take this approach because bad people broke the Internet and now you have to have a great honking opaque TLS layer between you and the caching servers so there's no way for this optimisation to work any more.
Of course it isn't really as important these days because we've got faster connections and interactions with the server are far less transactional and richer. But I still would like to have a way of tracking my own webusage and being able to go back in time without having to actually revisit each and every site.
These days you have to hack the browser because that's where your TLS endpoint emerges. Kaspersky tried this for their HTTP firewall application and there was ructions over that.
I'll defo take a look at this. Sounds just like what I've been looking for.
> Isn't this how the internet was supposed to work in the first place? I remember Netscape navigator having a 'go offline' icon in the corner.
Thinking back actually, if you forget about "the web"/HTTP - then yes actually - this is exactly how usenet worked and now I'm remembering that the "go offline" button used to download all your newsgroups along with your email and stuff so you could look at it all offline :-)
If you want something that's like Usenet these days check out Scuttlebut.
Nothing in the HTTP spec has changed about this AFAIK. The internet still behaves this way.
> Anyway, you can no longer take this approach because bad people broke the Internet and now you have to have a great honking opaque TLS layer between you and the caching servers so there's no way for this optimisation to work any more.
You could certainly still do this, you just need to import the proxy's CA certificate into your browser. It's just not possible to do to others on your network without consent now. (This is a security feature, not a bug.)
I was under the impression that TLS everywhere breaks the web as it was designed, ergo though nothing in the spec has changed - the environment itself has changed.
You could probably do something with isomorphic encryption but nobody seems that bothered. It’s probably of limited value nowadays as we use the web more as an rpc layer.
Setup will vary by proxy and OS but here's one example (which I haven't used but at a glance seems OK):
https://www.ssltrust.com.au/help/setup-guides/setup-squid-pr...
https://www.google.com/amp/s/amp.theguardian.com/world/2014/...
https://support.mozilla.org/en-US/questions/1263417
So if you have a local web server (like on localhost), you'll still be able to access it, but you won't be able to access non-local IPs.