LLMs have a a tendency to hallucinate: https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

As someone else stated, the AI can’t reason. It doesn’t understand what a unicorn is. It can’t think “a unicorn has a singular horn, so a non existent two-headed unicorn would have two horns”. Somewhere along the line it’ll probably mix in a deer or a moose that has two horns, because the number two matches the number of horns per head statistically.

Last year, two lawyers in separate cases with different LLMs submitted hallucinated case citations. It would have been trivially simple for them to drop the case number into a proper legal search engine, but neither did. This is a similar issue: the LLM will also prioritize what you want to hear, so it does what it’s designed to do and generate text related to your question. Like the unicorn example, it has no reasoning to say “any legal research should be confirmed by making a call to an actual legal database to confirm citations” like a human would. It’s just scribbling words on the page that look like other similar words it knows. It can make case notes look real as heck because it has seen other case notes, but that’s all it’s doing. (please excuse the political news story, but it’s relevant)

• https://www.npr.org/2023/12/30/1222273745/michael-cohen-ai-fake-legal-cases
• https://www.bbc.com/news/world-us-canada-65735769

And it’s not limited to unicorns or case notes. I found this reddit post while researching a feature of a software package (Nextcloud) several months ago. In the post, OP is seeking an option to pause the desktop client from the command line. Someone responds with a ChatGPT answer, which is quite hallucinated. Not only does such an option not appear in the documentation, there’s an open bug report to the software devs to request that the feature be added. Two things easy for a reasoning human to do, but the AI is just responding with what you want to hear - documentation.

• https://www.reddit.com/r/NextCloud/comments/136hm7j/pauseresume_sync_from_command_line/
• https://github.com/nextcloud/desktop/issues/5649

I’ve also seen ChatGPT tell my friend to use power shell commands that don’t exist, and he has to tell the model twice to generate something new because it kept coming to the same conclusion.

It’s a perspective trick - they’re not close as you think they are, and the United plane carries more than double the passengers of the closer Alaska.

1. Where is the server located? Are you looking at an intranet location or internet?

2. Is the client connected to the VPN concentrator via IPv4 or IPv6?

3. Is the VPN concentrator connected to the server via IPv4 or IPv6?

What you ask may be possible depending on those answers.

Going to summarize a lot of comments here with one - VPNs are very powerful tools that can do lots of things. Traffic can be configured to go in several directions. We really have to know more about your use case to advise you as to what config you might need.

Going to just write a ton of words on paper here - OP, let me know if any of this sounds like what you’re trying to do, and I can try to give a better explanation (or if something was confusing, let me know).

VPN that uses the client’s IP when sending data out of the VPN server

That’s the specific sentence I’m getting caught on myself. It could mean several things, some of which have been mentioned, some haven’t.

• Site to site VPN: Two (generally) fixed devices operate a VPN connection between them and utilize some form of non-NAT routing so that every child device behind each site sees it’s “real” counterpart without getting NATed. However, NAT is typically still configured for IPv4 facing the internet, so each device shows an internet “exit IP” matching the site it’s on. Typically, the device with the most powerful / most stable / most central / least restrictive would be the receiver, while the other nodes would be initiators pointed to that receiver. In larger maps, you could build multiple hub/spoke systems as needed.

• Sub-type of site to site possible: where one site tunnels all of its data over to the second site, and the second site is the one that provides NAT. This is similar in nature to how GL.Inet routers operate their VPN switch, but IMHO more powerful of you have greater control over the server compared to subscribing to a public VPN service. Notably for you example, the internet NAT exit device can be either the initiator or the receiver.

• Normal VPN but without NAT: this is another possible expansion of what you’ve written, with one word adjusted - it operates the VPN but preserves the client IP as it’s entering the network. This is how most corporate remote access VPNs operate, since it would be overloaded and pointless to have every remote worker from a small pool of IP addresses when you don’t even need to use a NAT engine for intranet.

My remote access VPN for my home lab is of the latter type, and I have a few of the sites to site connections floating around with various protocols.

For mine, I have two VPN servers: one internal server that works tightly with my home firewall, and one remote server running inside a VPS. Both the firewall and VPS apply NAT rules to egress traffic, but internal bound traffic is not NATed and simply passed along the site to site connections to wherever it needs to go. My home-side remote access VPN is simply a “dumb” VPN server that has the VPN protocol port forwarded back to it and passes almost raw traffic to the firewall for processing.

For routing, since each VPN requires its own subnet, I use FRR with a mixture of OSPF and iBGP (depending on how old the link is)

For VPN protocols, I currently am using strongSwan for IPsec, but it’s really easy to slap OpenVPN onto that routing stack I already set up and have the routes propagate inward.

Wasn’t there a subreddit for Cat Distribution System?

Even Opera is now Chrome…

Sometimes I’m watching something to analyze or I just really liked a song.

I also use j/l, left/right, and ./, for those three different types of scrubbing

The only number skip I use is 0 to rewind. Kinda helpful NGL.

Any VPN that terminates on the firewall (be it site to site or remote access / “road warrior”) may be affected, but not all will. Some VPN tech uses very efficient computations. Notably affected VPNs are OpenVPN and IPSec / StrongSwan.

If the VPN doesn’t terminate on the firewall, you’re in the clear. So even if your work provided an OpenVPN client to you that’s affected by AES-NI, because the tunnel runs between your work laptop and the work server, the firewall is not part of the encryption pipeline.

Another affected technology may be some (reverse) proxies and web servers. This would be software running on the firewall like haproxy, nginx, squid. See https://serverfault.com/a/729735 for one example. In this variation of the check, you’d be running one of these bits of software on the firewall itself and either exposing an internal service (such as Nextcloud) to the internet, or in the case of squid doing some HTTP/S filtering for a tightly locked down network. However, if you just port forwarded 443/TCP to your nextcloud server (as an example), your nextcloud server would be the one caring about the AES-NI decrypt/encrypt. Like VPN, it matters to the extent of where the AES decrypt/encrypt occurred.

Personally, I’d recommend you get AES-NI if you can. It makes running a personal VPN easier down the road if you think you might want to go that route. But if you know for sure you won’t need any of the tech I mentioned (including https web proxy on the firewall), you won’t miss it if it’s not there.

Edit: I don’t know what processors you’re looking at that are missing AES-NI, but I think you have to go to some really really old tech on x86 to be missing it. Those (especially if they’re AMD FX / Opteron from the Bulldozer/Piledriver era) may have other performance concerns. Specifically for those old AMD processors (Not Ryzen/Epyc), just hard pass if you need something that runs slightly fast. They’re just too inefficient.

Counterpoint: if you system is configured such that the mere act of trying to send an email results in serious delays and regular bounces, you’re doing email wrong. Even push notifications may require third party routing through Google, Apple, or similar to get to the core OS in some cases.

Yes, I recognize that hosting an SMTP server is difficult these days and can’t always be done at home due to IP restrictions. But that doesn’t mean you have to have an email server at home. I have a third party email on my domain and I can dispatch SMTP which arrives at expected non-delayed times even to Google and Microsoft accounts.

I honestly wish more software would simply speak to an SMTP server of choice rather than defaulting to just hitting the CLI mail send or attempting a direct SMTP connection.

Actually, I legally can’t make money off of it for reasons that would dox me.

I already pay for both VMware and Microsoft licensing among several others. If I can get my SSO by saving a little bit of money by using a different product, I will. I don’t mind paying for software I use when it makes sense, I only disagree with companies up-charging features like SSO that should be available to all customers.

You have my sword!

You’ll find a lot of pessimistic people here because there are few unicorns when a commercial company buying an open source project didn’t go badly for the open source people. Most of the time after a sell-out the projects ends up under highly restrictive licensing, features behind paywalls, and many other problems making it a shadow of its former self.

The most notable recent examples I can think of is IBM buys Red Hat buys CentOS, and that ended with forks as AlmaLinux and Rocky Linux. Oracle buys MySQL ended up forked as MariaDB. Businesses love to push their commercial offerings on open source products, and it’s not always in the form of plain old support agreements (like the people behind AlmaLinux). Often (this is common especially in databases) they’ll tax features like SSO, backups, or literally simple the privilege of having stable software. Projects like CentOS and VyOS don’t have stable OSS versions, and soooo many databases will put LDAP/Kerberos behind the commercial product, charging monthly or yearly operating costs.

Even GitHub (which to be clear was closed source to begin with, but is a haven for F/OSS so I’ll give it an honorable mention here) started showing Microsoft-isms after M$ bought the platform.

I have something like 40-60 machines between hypervisors, VM, and physical. Central auth is an absolute must for that scale. Sure I could just re use the same password 60 times, but if that gets popped, I’d also have to change it 60 times (adding config management is a soon to be completed task)

Just because you’ve used it professionally, doesn’t mean it’s OK.

Run the installation file to install the RDPwrap dynamic link library (DLL). This software provides the necessary functionality to enable Remote Desktop from a Windows 10 Home system.

  begin
    if not Reg.OpenKey('\SYSTEM\CurrentControlSet\Control\Terminal Server\Licensing Core', True) then
    begin
      Code := GetLastError;
      Writeln('[-] OpenKey error (code ', Code, ').');
      Halt(Code);
    end;
    try
      Reg.WriteBool('EnableConcurrentSessions', True);
    except
      Writeln('[-] WriteBool error.');
      Halt(ERROR_ACCESS_DENIED);
    end;
    Reg.CloseKey;

So essentially the RDPwrap software subverts Windows 10 Home security to enable Remote Desktop Connections.

Even without disassembling their shim DLL, just their readme language and installer code doesn’t give me warm fuzzies about this software’s ability to survive legal scrutiny or a Microsoft audit.

Just like with backups, in my professional IT Admin opinion: if its expensive enough to need remote access, it’s expensive enough to remote access the right way. There’s plenty of free remote options on Windows that don’t require monkey patching the core services and using a Home license professionally. Plus, if you have more than a few Windows installs, you probably want Group Policy anyways, so you’re up to the pro license key for that anyway, plus the Windows Server license key(s) for the AD controller.

Yeah, windows is expensive when used professionally. If you need windows that badly deal with it or talk to your software vendors about getting Linux or Mac software.

If you’re ok leaving a monitor plugged in (but can be off), my go-to is Parsec. Bonus points is that it works without needing a VPN (it uses UDP NAT hole punching like Chrome Remote Desktop). If you’ll be far far away from home, Chrome Remote Desktop tends to be slightly more reliable over high latency than parsec for me - but that could just be because I tuned mine for super low latency when nearby.

Good news is, you can run both at the same time and see how they treat ya! (And both are free for base use, but parsec has a handful of premium features you can pay for if you like it) I have Parsec, CRD, RDP, and SSH all set up in various forms to get back “home” when I’m not.

(if this comment reads like I feel slighted it’s because I do)

Their networking ecosystem is very focused on a specific class of prosumer and once in it can be very difficult to upgrade out of that bubble to toys that have more growth capacity, from both a tech and learning perspective.

I have an advanced network with dynamic routing (iBGP and OSPF), as well as several VPN protocols for both site to site and access VPN. I also have redundant layer 3 gateways everywhere in the main site. Ubiquiti has had the tech to make redundant layer 3 for YEARS, but they refuse to and instead stop updating useful product lines that have more features and instead focus on gimmick products that have flashy marketing campaigns. Even on one of their more feature-ful routers (ER-4), I have to use OpenVPN gateway servers because Ubiquiti doesn’t support plugins that I can get on *sense for full mesh VPNs.

I can really only use them at layer 2 because once I hit my network core I need redundancy protocols at L2 (stacking or vPC/MLAG) to maintain a system that can keep vSAN and Ceph happy.

I’m really glad I went the *sense route instead of taking a chance on a USG-3 and depending on the custom json file to load OSPF, because that’s a feature they removed from newer gateways iirc.

Can’t tell you how many times I’ve googled things and found my own posts and bug reports.

Voyager shows those both as strikethrough

This elicited a genuine laugh from me.

Excellent work, OP. I can feel the scope creep in my bones.