Hacker News Reader

At some point, internet connections will be fast enough to make them indistinguishable form a local machine. Today, that is not true.

In my own job, I straddle cloud-based and local machine-based software. Overwhelmingly, the cloud-based software is slow to the extent that it removes at least 40 points of IQ and offsets a decade of experience. The intellect and mental speed I have spent a lifetime building are simply washed away by latency that shuts my mind down, effectively turning me into a ninth-grader. Sure the company pays me as if I continue to generate output becoming of my generous salary. But I maintain the illusion simply by spending many hours on weekends with copious amounts of coffee and, at times, gin, doing work on cloud-based software that, on the "outdated" local software, used to take minutes.

Cloud software is among the biggest frauds ever perpetrated on the American consumer.

I play PC and VR games on ShadowPC. I can’t tell the difference with a controller, and the latency with a keyboard and mouse is like a cheap Bluetooth mouse. I think the key for these is relatively local data centers.

Once ISPs cohost, or get in on the game themselves, we’re set!

Having the database on the same server as the application significantly improves responsiveness of applications.

Many of these cloud applications choose fragmentary architectures with data sprinkled across many discreet servers that are milliseconds or tens of milliseconds away from each other, resulting in significant delays compared to doing these blocking operations against a local database.

Cloud native isn't always the right choice, sometimes CRUD apps should have a local database just for the responsiveness benefits.

Having databases on the same server as the application can also slow both down and introduce potential security issues.

> At some point, internet connections will be fast enough to make them indistinguishable form a local machine

Has latency really improved that much lately? At least for gaming latency has been more or less constant for me over the last ~10 years or so.

And if the server is in the same city or quite close to you I think it’s already possible but possibly not that useful.

This here. Everyone talks like bandwidth is what's lacking. The reality is latency is the problem and you can't overcome the physics (currently). MS's cloud PC's can get stuffed.

> This is how IT worked 40+ years ago:

It is also how it worked 20-10 years ago: it was called Terminal Services.

And then it was neglected because next shiny thing: VDI.

The thing is, one big server for a hundred office users doesn't cost much, because they are sharing the RAM. But with VDI you need at least 100 x 4Gb memory and that means you need way more than one server. And some SAN, because only SAN with fast drive can survive the morning boot storm. Oh, by the way, here are two FC SAN switches for your HA storage fabric, because you really want paths to the SAN be redundant.

But anyway, local installs are not bringing revenue constantly. But renting the services do. This was perfectly clear with AzureAD and Exchange 365.

Not dead yet, my consulting business is almost entirely built around provisioning terminal servers as small private cloud managed systems for companies with niche business requirements. Not sure how much longer the model is going to work because eventually I suspect Microsoft will simply kill off on-prem terminal services all together or make it completely uncompetitive with their own azure / Office 365 combination.

But the surfeit of stupendously powerful white box server platforms and the remarkably efficient way that a terminal server conserves memory usage across sessions compared to containers or vdi or anything else, absolutely outperforms almost every other solution in the SMB space, in terms of how each support engineer can scale to 100 or more end users across a half dozen completely unrelated organizations. I'm surprised how underrepresented this business model is these days.

> I'm surprised how underrepresented this business model is these days.

VARs are living on %% of the sale.

> Not dead yet

Yep, but only yet.

This cycle of alternating didn’t start 40 years ago, which would be 1983. We have been cycling between decentralization and centralization since at least the 1970’s.

Because they both suck, either side “winning” is worse than staying in the middle. But we only like pendulum swinging in this industry so nobody stops as the bottom of the arc.

> dumb terminals that would turn into doorstops if the network connection with the server failed for any reason.

OTOH, you could log off from one terminal and go to any other terminal (maybe in a different country) and continue to work right where you left.

If the environment is designed for that, it's not only doable, but practical.

I'd personally rather achieve this by remoting into a machine that I already own, rather than paying rent to Microsoft in perpetuity.

True, but aren't there far more economical ways of doing it if you're an IT department? Running a full Windows VM with 4G to 8G of reserved memory seems incredibly wasteful if all you want to do is centralise the provisioning and management software and data.

Why not Chromebooks + web apps if you need some client-side hardware anyway?

> Why not Chromebooks + web apps if you need some client-side hardware anyway?

Some companies are married to Microsoft Office, Exchange, and ShamePoint.

That's true and one great advantage. But I don't get how to apply this advantage to personal computing. I have a laptop and I never ever wanted to randomly "log into my cloud workstation" from some other place. I can just open the lid of my laptop and continue working. What advantage does the requirement of a persistent network connection offer me? Seems like working on a train is not doable.

I could see a use case where people don't want / can spend to much money on hardware. But even then, hardware is still required and I can't see where a subscription would help here.

One good reason is to prevent data exfiltration, or, at least, make it less convenient.

That does only apply to business environments, but my comment was regarding personal computing.

It's still harder if you can't run arbitrary code on the remote computer or transfer the files there to your local one.

I run almost all of my devices as “dumb” terminals for most purposes.

I can work in my office at my workstation. I can walk out to the kitchen and sit down with my laptop and open the lid and be _exactly_ where I just was. I can move across the country and work from my laptop, a friend’s computer, a library, or anything else I run across along the way. My house can burn down and I can go to an electronics store and buy a new laptop and continue typing in the middle of the line where I left off when the smoke detector started beeping and I ran out of the house.

Separating the device I’m using from the work I’m doing makes the friction to move between devices near zero or zero.

I never use my friend's computers, library computers, or anyone else's computers to log into my personal accounts when I can afford to have my own devices. The thought of entering passwords into devices that you don't own and trust horrifies me. Even if someone is trustworthy, their devices could be compromised. Not to mention that it can make your friends uncomfortable if you're borrowing their computers all the time.

It's not like I'm running around living my life on other people's computers. The guy I replied to said he had no idea how to apply this to personal computing to get any advantages. "I spend 90% of my time sitting at my desk, using my desktop, doing normal boring work." isn't really a good demonstration of any advantage.

Most of my work _is_ done from my desktop. The bulk of the remainder is done from my laptop. For that alone I find enough advantage to keep doing it. Zero context lost bouncing around.

But when I _was_ moving across the country, it was certainly nice not having my ability to earn an income any given day tied to a single device. When I was effectively homeless in my month-long move and staying with family at one point I did end up using their computer a couple times. You'll just have to take my word for it that they were okay with it.

The assumption you make is that all these devices you use have a good internet connection. This isn't always true for a lot of people.

I'm not making any assumption. I'm speaking from my personal experience.

Yes, if I'm somewhere without internet I can't get a lot done. But that's true whether I'm running programs locally or remotely.

Based on the number of people that show up on HN every time GitHub/GitLab/etc go down, I don't think I have any unusual dependencies that nobody else does.

During my summer trainship, accounting was run on an AS/400 used by dumb terminals on the harbour offices.

At the university we had DG/UX, Solaris and eventually Red-Hat Linux, with a mix of classical phosphor and green terminals, PC and classical Macs (LC models).

At Nokia and CERN we used to develop on UNIX (HP-UX, Solaris, Scientific Linux) servers, PC and Macs were only used for Office and as thin clients for the R&D servers.

And then we moved into Citrix, RDP and Terminal Services for Windows development, so it has been with us all along.

The centralization to decentralization pendulum swings again