[kevingadd]

Large downloads dramatically decrease the number of users who get from a download page to actually running an application, even for web browsers. As a result, any steps you can take to make an install process faster or easier will have a tangible impact on the number of customers you have (and any revenue potential that results). Mozilla, like any other company, has salaries to pay, so they would be foolish to make things worse for the majority of users just to make them slightly easier for theoretical users who want to install in a networkless VM and can't figure out that the stub is a stub.

It's a complication, but it eliminates the need for the user to come back after waiting for some dozens of megabytes to download and then interact. You can download a stub near-instantaneously, do things like pick your install directory, and then forget about the install process entirely as it completes without your supervision. That's a good thing, and it isn't possible with offline installers.

Browsers like Firefox and Chrome already need network access after installation anyway. They update on a regular basis, and pull down optional packages like Chrome's software WebGL rasterizer, malware blacklists, and Firefox's GPU blacklist. So, in practice, it is already impossible to do a true offline install of either browser; you just may not be aware of the things that are left out of the installer.

[qwerty_asdf]

  > Large downloads dramatically decrease the number of 
  > users who get from a download page to actually running 
  > an application

I think I almost entirely agree with you on that.

  > they would be foolish to make things worse for the 
  > majority of users 

I'm arguing that this provision (clearly pointing the way toward an offline install) would not make things worse for anyone.

  > just to make them slightly easier for theoretical users

I am not a "theoretical" user. I am an actual user.

  > ...do things like pick your install directory, and then 
  > forget about the install process entirely as it 
  > completes without your supervision. That's a good thing, 
  > and it isn't possible with offline installers.

What?! What are you even talking about? I can pick my install directory with the full installer. I double click the icon, it asks me for the directory, it runs, and then the end.

Firefox's installer is dead simple, and believe me, I am HUGELY grateful for that. The only thing easier than the Windows installer is unzipping the Linux tarball into whatever path I choose. I hate Adobe's Flash and PDF installers, and I fucking DESPISE the Java installer, with its damnable bundling of that bloody Ask.com toolbar.

Seeing Firefox add these additional layers of background updater services and stub installers deeply worries me, and fills me with concern that as a "salary paying organization" (albeit, ostensibly non-profit), they might veer down an ugly path, go the way of the Sith, and start engaging in questionable behavior that is inappropriate for an open source project.

Consider the example of Ubuntu's desktop file search bundling Amazon ads in the results. What if one day, Mozilla decides that in order to pay the bills, it needs to negotiate a deal, whereby all those users with automatic updating enabled should get railroaded with some kind of optional-but-defaulted-to-enabled third party feature suggesting helpful reminders to buy more burritos from Jimbo's Refried Beans Emporium. Just sayin'...

  > Browsers like Firefox and Chrome already need network 
  > access after installation anyway. 

Not to do things without my permission, they don't. This "need" business... I disagree.

A browser only "needs" to to exactly what I ask it to do, and not much else. I tell the browser what to do, not the other way around.

  > They update on a regular basis, and pull down optional 
  

Optional. That's an important adjective in your sentence.

  > packages like Chrome's software WebGL rasterizer, 
  > malware blacklists, and Firefox's GPU blacklist. 

Yeah, the very same blacklist that I'm frequently bypassing to check out all the cool Web GL experiments people post here on HN. I know all about that.

Malware blacklists are another concept that I tend to reject as mostly ineffective in achieving their stated goal. We could go round and around with that argument for days. Let's not get started on THAT can of worms.

  > So, in practice, it is already impossible to do a true 
  > offline install of either browser; you just may not be 
  > aware of the things that are left out of the installer.

When you use the word "impossible", I have to just flatly disagree with you. And in general, most of the things that you mentioned are things that I don't care about, and will never be interested in.

[kevingadd]

As I explained in the parent post (which you failed to parse correctly - was my English too complex? I'm not the greatest at writing clearly), the advantage of the new streaming installer is that all user interaction is frontloaded, not that it adds new UI. I don't know why you thought I was claiming the ability to pick an install directory is a new feature.

Anyway, my point re things like the blacklist and webgl rasterizer is that the browser is not completely installed without those components. The offline installer is missing key features and components of the browser; things that show up on feature lists that web applications can rely on. Without the WebGL rasterizer, WebGL demos won't work in your offline-installed copy of Chrome in your VM without hardware accel. Without the GPU blacklist, on some configurations Firefox will be nonfunctional (or worse, crash your machine) because it attempts to do acceleration on a broken driver. Audio/video codecs are another area where it's no longer possible to realistically ship 'everything' in a single offline installer.

If you simply want an offline installer for a bare minimum browser that can do a stripped down subset of HTML5, it's still possible to deliver that. But the amount of the 'web' that works in that bare minimum installation will keep decreasing.

P.S. The reason Firefox has to install a service is because it's not possible to install updates cleanly in any other fashion on UAC-enabled Windows. Your alternatives are installing into %AppData% (like chrome does, which removes the ability to pick an install dir and has other gross consequences) or requiring the user to UAC elevate every time an update installs. As I stated in the parent post, not installing updates puts users at risk when you're dealing with a web browser - the attack surface is enormous.

[qwerty_asdf]

  > The reason Firefox has to install a service is because 
  > it's not possible to install updates cleanly in any 
  > other fashion on UAC-enabled Windows.

Ah ha! Now you're speaking my language! That I understand perfectly!

It makes sense that Windows UAC has forced bad decision making. Windows, in general, is just a hideous, mutated mess nowadays.

Anyway, it's essentially a correct decision to never marry a browser to device drivers or hardware. The websites that count won't dare crash browsers due to hardware requirements. The browsers that count will fail gracefully, and tell the user that their settings are not compliant with the requirements of the page they are trying to view, before they ever crash. However one wishes to communicate the nature of highly specific user options, and then adapt to the long fragmented tail of hardware conditions, is beyond the scope of a "NORMAL" web browser (no-true-scotsman).

When requirements become that unforgiving, you've entered the realm of the highly specialized plug-in, or custom client-server software. It's cool that Firefox is brave enough to wade into those territories, and still deliver awesomeness, but the core necessities of the web browser should never be sacrificed, for fluff and sugar-coated eye candy.

I'm really not worried about the idea that "The Web" is changing. The sites that matter will always work with the bare minimum, with little more than HTTP GET and HTTP POST, even with JavaScript disabled.

The rest is just cruft.

[baddox]

All your points are valid, but none of them explain why Mozilla couldn't or shouldn't just put an explicit link from the main download page to something like "offline install" or "redistributable install."

[DanBC]

> So, in practice, it is already impossible to do a true offline install of either browser; you just may not be aware of the things that are left out of the installer

Some people have lousy Internet connections. Those people use things like download managers to help. Download managers don't work with stub installers.

A simple link to a true offline installer would help those people. (Torrent is fine too, I guess.)