Unfortunately the reality of IPFS is that despite its huge funding it was poorly designed from the start and still to this day has much slower loading times then my I2pd instance (despite i2p transmiting messages through multiple encrypted proxies), to the point where the company working on the rust implementation determined it was so bad they had to scrap the whole thing to make something that actually worked. Not to mention that I managed to have my server taken over by some kind of malware by downloading a particular piece of content.
Thanks, that was an interesting read! I always felt IPFS wasn't ready yet, but the value it tries to provide of being a file system, I've found no real alternative to.
Very good to read that iroh is willing to look beyond the IPFS spec to provide its values with better performance. I hope it works out.
Also the user interface and builtin solutions for torrenting, hosting, address booking make it way more user friendly for people to start using I find.
Technically yes by rewriting ipfs's code, but due to ipfs's flaws you would be better off using something like freenet/hyphanet which has been designed for that purpose and has been successfully running since 2000, with the added benefit that the data is actually stored in the network by others instead of just by you (at least when you often request the data)
OpenTelemetry and in particular I wish more protocols had Traceparent propagation support and more software had support for sending spans and traces to an OTLP endpoint to construct a full picture of everything that is going on in a distributed system.
There are a bunch of message broker services out there, and having a consistent set of common keys along with a documented process for transforming events to/from different systems means that this kind of data can move through different systems without getting mangled. It does have a spec for JSON, so it can be considered just a standardized JSON blob with transformation rules. But it also has a protobuf spec, specs for MQTT, NATS, HTTP, Avro, etc. It’s a common language for all these systems.
I wish Microsoft Office would use the .odf standard by default. Or, failing that, it'd implement its own published .docx specification correctly, so other office suites can be compatible.
The entire purpose of Microsoft standardizing OOXML and implementing it wrongly in Office was to make other office suites irrelevant. ODF was already standardized and countries would have adopted it if MS didn't do the same with OOXML. They stuffed the ISO with members supporting them to do it.
And now that OOXML is a viable standard, they implement it wrongly so that other office suites can't be compatible with MS Office without a lot of extra effort. Any incompatibilities with MS Office will be considered as the fault of other office suites by the general public and government officials.
Expecting MS to do what's right for the customers is putting too much faith in their nonexistent sense of ethics.
Markdown. Its only in tech-spaces that its preferred, but it should be used everywhere. You can even write full books and academic papers in markdown (maybe with only a few extensions like latex / mathjax).
Instead, in a lot of fields, people are passing around variants of microsoft word documents with weird formatting and no standardization around headings, quotes, and comments.
My main wishlist for markdown, is a better live collaborative markdown editor. Hedgedoc works, but it's showing it's age, and they don't seem to be getting close to releasing v2.
Etherpad also has a markdown extension, but it doesn't import / export that well.
Markdown is awesome, I agree! I did not realize you could extend markdown with anything other than html. The html extension is quite nice to do anything that markdown doesn't support natively, but I wish there was an easier way to extend markdown. Maybe the ones you listed are what I need.
Markdown is terrible as a standard because every parser works differently and when you try to standardize it (CommonMark, etc.), you find out that there are a bajillion edge cases, leading to an extremely bloated specification.
Commonmark leaves some stuff like tables unspecified. That creates the need for another layer like GFM or mistletoe. Standardization is not a strong point for markdown.
I believe commonmark tries to specify a minimum baseline spec, and doesn't try to to expand beyond that. It can be frustrating bc we'd like to see tables, superscripts, spoilers, and other things standardized, but I can see why they'd want to keep things minimal.
Asciidoc is a good example of why everything should be standardized. While markdown has multiple implementations, any document is tied to just one implementation. Asciidoc has just one implementation. But when the standard is ready, you should be able to switch implementations seamlessly.
What's the alternative? We either have everything specified well, or we'll have a million slightly incompatible implementations. I'll take the big specification. At least it's not HTML5.
Depends on the type of book. Since you need HTML for all non default styles. Therefore, it raises the bar... you need a bit of web dev knowledge which removes the biggest benefit of markdown: simplicity / ease of use.
Man, I've written three novels plus assorted shorter form stories in markdown.
There's a learning curve, but once you get going, it's so fluid. The problem is that when it comes time to format for release, you have to convert to something else, and not every word processor can handle markdown. It's extra work, but worth it, imo.
Publishers are pissy about such things. Even self publishing (which is what I do now), the various outlets still have limits to what they will use. Amazon accepts something like three file formats, including their own, and pdf isn't on the list.
I could just do pdf for directly giving them away to people, but even then, epub is usually a better pick in terms of readability since that's the standard for actual books since ereaders tend to display it better than pdfs. Most people reading books via files would be using something that can give a better experience with epub vs pdf.
Pandoc + [your markdown editor of choice] is magic. Some editors even come with Pandoc as a dependency so you can export to more or less anything from the GUI. I think GhostWriter and Zettlr at least (I honestly can't be sure, I've changed editors so often and now I just have some Pandoc conversion scripts in my file manager menu).
ReST (restructured text) is a good middle ground. I just wish it had more support outside of the python community. It could use some new/better tooling than Sphinx
IOT devices shouldn't connect to wifi. ZWave or zigbee is much better suited to IOT stuff, but it seems to mostly get adopted in very limited, locked down proprietary shit like Hue Lights.
Yes but at least Hue (and IKEA and LIDL and many other brands') lights work well with open Zigbee coordinators, like deconz and ZHA in Home Assistant.
I wish there were more Zigbee and Zwave and less WiFi IoT devices too. I don't even have a Zwave coordinator because I never found anything I wanted with Zwave support.
There's only one case I've found where Wi-Fi use seems acceptable in IoT: ESPHome. It's open-source firmware for microcontrollers that makes DIY IoT sensors and controls accessible over LAN without phoning home to whatever remote server, without trying to make anything accessible over the Internet, and without breaking in any way if the device has no route to the Internet.
I still wouldn't call Wi-Fi use ideal even there; mesh can help in larger homes and Z-Wave/Zigbee radios tend to be more power efficient, though ESP32 isn't exactly suited for a battery-powered device that's expected to run 24/7 regardless.
Rcs was designed from the ground up to be handled by the carrier in clear text like sms, it doesn’t incorporate encryption in any way and doesn’t do much at all to address the untrustworthy nature of carriers and law enforcement nowadays.
It’s like those two protocols started developing at the same time and only google kept extending rcs to keep some degree of feature parity with imessage.
If we had to ditch iMessage it ought to be for some third type, not for questionably secure rcs and what new bubble color can be used to indicate that someone’s using an unencrypted rcs server?
Google has used RCS as their latest attempt at entering the messenger market. I really don't see why anybody else would want to adopt it under these circumstances. I mean Samsung did but Samsung is playing their own little paranoid game with Google, they don't really give a crap about RCS.
Basically Google killed RCS. They will never be able to make anybody adopt it against their will in the EU, people will stick to what messenger services they're already using. If they ever attempt to turn it on by default in their own app it will turn into a regulatory issue so fast their head will spin.
I want us to stop using communication protocols that are tied to our connectivity providers. Let alone tied to a specific piece of hardware (SIM card).
"Telephone providers" should be just another ISP. And whatever I do over the network shouldn't care if it is running on a mobile network or a landline fibre.
While we are at it let's fuck off with this SIM shit. You don't get to run code on my device. Give me an authentication key that I can use to connect to your network and then just transfer my packets. My device runs my code.
IPv6. Lack of IPv4 addresses it's a problem, specially in poorer countries. But still lots of servers and ISPs don't support it natively. And what is worse. Lots of sysadmins don't want to learn it.
If computers connect to others through the internet, the IPv6 address can reveal how many computers there are on the local network, and if certain traffic to different destinations are coming from the same computer, but also if one of the computers has gone offline but then resumes from sleep/hibernation.
To me their comment means they want to avoid that, and I agree, I want to avoid that too. To fix these, I would need to configure NAT on my router for IPv6.
Yes IPv6 address privacy extensions help somewhat, but
computers won't use a different v6 address for every distinct destination, they will just start using a new one from time to time
computers won't stop using the old v6 address immediately after wakeup
With v4 addresses these did not really matter, because everything was being sent from the same public IP, and and outside observer could only see what a "network" is doing collectively. But with v6 an address identifies a computer, across websites/services. Even if it's just for a "short' time, even if the address is randomized.
If you want privacy, you need some kind of VPN or onion routing. Even if everything you list were correct, the difference between IPv4 and 6 for privacy would be marginal.
I don't think this is so black and white. I'm a regular tor user, but so often it's not worth it to load webpages through a dial-up connection, and then there are the sites that block access for tor users for some reason.
Even if everything you list were correct
Which parts weren't?
the difference between IPv4 and 6 for privacy would be marginal
You're thinking of a firewall. NAT is just the thing that makes a connection appear to come from an IP on the internet when it's really coming from your router, and it's not needed with IPv6. But you would not see any difference with IPv6 without it.
You're thinking of a firewall. NAT is just the thing that makes a connection appear to come from...
That connection only "appears to come from" if I explicitly put a rule in my NAT table directing it to my computer behind the router doing the NAT-ing.
Otherwise all connections through NAT are started from internal->external network requests and the state table in NAT keeps track of which internal IP is talking to which external IP and directs traffic as necessary.
So OP is correct, it does apply a measure of security. Port scanning someone behind NAT isn't possible, you just end up port scanning their crappy NAT router provided by their ISP unless they have specifically opened up some ports and directed them to their internal IP address.
Compare this to IPV6 where you get a slice of the public address space to place your devices in and they are all directly addressable. In that case your crappy ISP router also is a "proper" firewall. Strangely enough it usually is a "stateful" firewall with default deny-all rules that tracks network connections and looks and performs almost exactly like the NAT version, just without address translation.
So OP is correct, it does apply a measure of security. Port scanning someone behind NAT isn't possible, you just end up port scanning their crappy NAT router provided by their ISP unless they have specifically opened up some ports and directed them to their internal IP address.
You end up just port scanning their crappy router on IPv6 as well because ports that are not opened are stuck at the firewall either way, no matter if you use IPv4 or IPv6.
Just because every device gets a public IP does not mean that IP is publicly accessible.
An advantage that IPv6 has against port scanning is the absurdly large network sizes. For example, my ISP gives me a /56 prefix, that is 4,722,366,482,869,645,213,696 IPv6 addresses. Good luck finding the used ones with the port open you need.
Even with just a /64 prefix you get 18,446,744,073,709,551,616 addresses, way outside the feasibility of port scanning.
Compare this to IPV6 where you get a slice of the public address space to place your devices in and they are all directly addressable. In that case your crappy ISP router also is a “proper” firewall. Strangely enough it usually is a “stateful” firewall with default deny-all rules that tracks network connections and looks and performs almost exactly like the NAT version, just without address translation.
realistically, it wouldnt surprise me if ISPs started NATing on residential IPV6 networks, just for the simplicity, but still allowed end users to assign their own IPs if they so pleased. Given the surge in shitty IOT devices, that's probably a good thing for most people. Though a firewall would also accomplish this as well.
My isp decided to put me behind a CGNAT and broke my access to my network from outside my network. Wanted to charge me $5 a month to get around it. It's not easy to get around for a layman, but possible. More than anything it just pissed me off that I'd have to pay for something that 1 day ago was free.
Set up a reverse proxy on another machine (like one of those free oracle cloud things). I can't go into detail because I don't know exactly how. I think cloudflare also has options for that for free. Either way it's annoying.
Lots of really large sites are horribly misconfigured. I had intermittent issues because one of the edge hosts in Netflix ‘s round robin dns did not do MTU discovery properly.
My university recently had Internet problems, where the DHCP only leased Out ipv6 addresses. For two days, we could all see which sites implemented ipv6 and which didn't.
Many big corpo sites like GitHub or discord Apperently don't. Small stuff like my personal website or https://suikagame.com do.
Say this to my very large Canadian ISP who still doesn’t support IPv6 for residential customers. Last I checked, adoption in Canada was still under 50%.
50%?? I fucking wish. In Spain we are at 5%. I finally got IPv6 in my phone this year, but I want it in my home, which is still only available as IPv4 even if they're the same ISP.
I wish people used email for chat more. SMTP is actually a pretty great protocol for real time communication. People think of it as this old slow protocol, but that’s mostly because the big email providers make it slow. Gmail, by default, waits ten seconds before it even tries to send your message to the recipient’s server. And even then, most of them do a ridiculous amount of processing on your messages that it usually takes several seconds from the time it receives a message to the time it shows up in your account.
There’s a project called Delta Chat that makes email look and act like a chat app. If you have a competent email service, I think it’s better than texting. It doesn’t stomp on the images you send like SMS and Facebook do, everyone has it unlike all the proprietary services, and you can run your own server for it that interacts with everyone else’s servers.
Unfortunately, Google, Microsoft, etc all block you if you try to run your own server “to protect against spam”. Really, I’m convinced that’s just anticompetitive behavior. The fewer players are allowed to enter the email market, the less competition Gmail and Outlook will have.
As much as I like ProtonMail too, unfortunately their encryption models prevents it from working with Delta Chat. I’d love to see Proton make a compatible chat app that works with their service.
I made an email service called Port87 that I’m working on making compatible with Delta chat too. I’d love to see people using email the way it was originally meant to be used, to talk to each other, without being controlled by big businesses.
Oh, another awesome thing about email is that you can ensure that your address is always yours, even if you use an email service provider like Gmail. Any provider that supports custom domains will allow you to use your own domain for your address, then if you want to change your provider, you keep your address. So, since I own hperrin.com, I can use the address me@hperrin.com, and I know it’ll always be mine as long as I pay for that domain.
This is a much better model than anything else. Even on the fediverse, you can’t have your own address unless you run your own instance.
If your email service provider goes out of business or gets sold off (skiff.com, anyone?), as long as you’re on your own custom domain, your address is still yours.
I’m working on custom domains for Port87. It’s definitely a feature I think every email provider should offer.
Yes, I shifted to my own domain after my default ISP of 20 years decided that email was just too hard, you know? They didn't outright say it, they just started batch processing emails so that I'd get all my daily emails at around 2 am the next day. Super handy for time limited password reset emails!
A few hours reading a guide and setting up a $5/mo linode email server with SPF and dmarc, a few more hours transferring 20 years of IMAP mail from my old account to a folder, and a month or so of changing a few site contact emails over each day when they emailed something to my old account, and now I've got an email server on my own domain that is 10 times faster at sending/receiving mail than my old ISP ever was.
And now I can have amazon@mydomain.com and career@mydomain.com and random other disposable addresses so that when they are inevitably sold off for the $$$ I can just dump them and maintain a spam free inbox.
greylisting will typically only be applied to people who you haven't interacted before, so I don't think it is a big deal. It would be similar to how many major chat apps hide away suspicious messages from new people in some "invites" section that is often hidden by default.
If you’re relying on the remote server to delete something, you can’t trust it no matter what protocol you’re using.
For a regular email, the chance to undo might be fine, but for real time communication, it’s just an unnecessary road block.
Maybe if it was optional per recipient, or per conversation, or better yet, depending on the presence of a header, it might be fine. Gmail only supports all-on or all-off.
If you’re relying on the remote server to delete something, you can’t trust it no matter what protocol you’re using.
I mean yeah I wouldn't bet my life on it, but for the 99% of regular communication it's fine. That's no reason to not have it in the protocol and muck around with 10 second delays instead.
SMTP is a terrible protocol. Text based for sending effectively binary data with complex header wrapping and "generate a random delimiter" framing. We really need a HTTP/2 of SMTP.
That being said I agree that it exists and works. The biggest blocker to more IM-style communication is largely the UI and user expectations. I have no problem having quick back-and-forths over email but most people don't expect it.
Honestly my biggest complaint is header wrapping. Technically you need to wrap lines at 998 bytes (not that any reasonable server actually cares). But in order to wrap a header you need to add spaces (because you can only break a line after whitespace). But where spaces are unimportant depends on each specific header. So you need to have custom wrapping rules for each header.
In practice no one does this. They just hope that headers naturally have spaces or break them in random locations (corrupting them) because the protocol was too stupid.
Binary protocols are just so much simpler. Give the length, then the data. Problem solved. Maybe we could even use a standard format for structured headers. But that would be harder to do while maintaining backwards compatibility.