TDLib – Build Your Own Telegram Client | Hacker News Reader

TDLib – Build Your Own Telegram Client

telegram.org

205 pointsTiredOfLife97 comments

Unless they release the server code, no, you can't build your own "Telegram" yet. The original article's title was misleading, thankfully the HN submission title was amended to correct it. Still useful of course for anyone creating third-party clients, but an entirely different matter.

Aye aye! Both Telegram and Signal scare me.

We're building an MIT/ZLIB/Apache2 Open Source End-to-End Encrypted alternative that is a fully P2P decentralized app (dApp), see: https://hackernoon.com/so-you-want-to-build-a-p2p-twitter-wi... !

Currently it supports signed/verified messages, haven't added cyphered/private messages yet to the API, but it is available in the SEA (Security, Encryption, Authorization) utility library underneath - but that should be coming soon, and of course love any contributions!

Also, the "server" has no special application logic on it, it is just another peer in the network that can help with WebRTC signaling or being a websocket relay. It cannot middleman the data, because every message is signed via the "client" (technically there are no server/clients, everybody is just a peer) using latest native Web Crypto API!

Here is a demo of an example social network app that is syncing character-by-character with cryptographically verifiable sources across a test network of real world devices: https://youtu.be/C3akdQJs55E

Again, all truly Open Source (ie, MIT/ZLIB/Apache2)!

Cool!

> So You Want to Build a P2P Twitter with E2E Encryption?

Saw this and immediately wondered: is this just twitterish features (primarily broadcast) or are there plans for private and group chats?

Tox is also worth checking out. Decentralized, supports video calls, and you're automagically given an ID without having to provide any personal data. https://tox.chat

Why does Signal scare you? All of their code is open source[1] and has had multiple public audits[2].

[1]: https://github.com/signalapp

[2]: https://eprint.iacr.org/2016/1013.pdf

It requires you to provide a phone number (read disclose your identity) to use it. And its homepage has the word "privacy" repeated several times. It looks like a poor joke.

You can easily use a VoIP number to set up Signal if that's really an issue for you.

The alternative (using Telegram's homegrown and dubious cryptography and lack of actually encrypted group messages) is significantly worse. Here's a nice summary of the issues: https://security.stackexchange.com/a/49802/43733

Telegram also requires a phone number so it is not any better than Signal. Regarding cryptography, anything is good as long as three-letter agencies cannot break it.

VoIP is easily traceable too. Pay with a credit card for the number? Use Google Voice? Bam, they’ve got you.

Yet the developer is world famous for his security work, not for something like inventing Facebook for Russians. Are you sure you can handle 100% anonymous connection to servers of apps that could at any time identify you based on your IP?

I am not trying to discuss who is the developer, I just wanted to say that you cannot require a phone number and talk about privacy at the same time. Phone number discloses your location and identity.

IP address is much easier to hide, for example by using a VPN.

In my opinion your top priority should be to make every message e2e encrypted between sender and recipient. Else your idea is just as useless as telegram

The client-server protocol is published. What stops you from writing your own server implementation?

There may be political motive to this.

Why is Telegram X issued by "Telegram Messenger LLP" while the original Telegram is issued by "Telegram LLC"?

Telegram LLC, which issues the original Telegram app on the App Store has been undergoing litigation with regards to its ownership. It is based out of Russia and a company called UCP with Krelim ties who 'bought' VKontakte has been suing Durov for full Telegram ownership pursuant to VKontakte 'purchase'.

Telegram Messenger LLP however is based in London and seemingly wholly owned by Durov crew. In 2014, Telegram Messenger LLP launched Telegram HD, a separate app from Telegram-LLC-issued-Telegram. Now it is launching Telegram X , another separate app (which was until very recently Challegram, an open source Telegram client, winner of Telegram contest, and purchased by Durov) .

Draw your own conclusions.

EDITED for clarity and details

https://themoscowtimes.com/articles/the-telegram-lawsuits-ex...

http://www.ewdn.com/2014/07/29/pavel-durov-has-cloned-telegr...

https://rusletter.com/articles/ucp_requires_to_recognize_tel...

http://www.frandroid.com/android/applications/securite-appli...

> Challegram, an open source Telegram client

Is this true? I can't find the source anywhere. Not even a dated one, if they had re-licensed it.

I like how they do very good engineering (just straight C++, Swift, with little to no external dependencies) with a very small team of apparently very talented people. Most of their products are typically owned by one or two people. It's also pretty smart of them to make all of the clients available as open source and keep the server implementations private. This way they don't have to license QT for example.

QT was very liberally licensed for a while. Are they back to copyleft or something now?

It started moving from LGPLv2.1 to LGPLv3 in Qt 5.4 with new features, and is now the norm since Qt 5.7 (latest in 5.10):

http://blog.qt.io/blog/2014/08/20/adding-lgpl-v3-to-qt/

https://www1.qt.io/faq/#_Toc_1_2

Behold, 12283 lines of very good engineering:

https://github.com/DrKLO/Telegram/blob/e9e40cb13ea942b148b25...

Takes a great engineer to make that work as well as it does..

Well the c++ code is pretty good... In any case it works much better than the signal app on android.

I've never had any problems with Signal for Android. With Telegram, though, I once woke up to find myself added to what looked like the first 500 groups in their database, alphabetically sorted. My phone was buzzing continuously from the stream of notifications. Had to remove myself from them one by one.

Looks to be UI related stuff, some of which looks automatically generated to some extent, some parts look boilerplate. Having files with many LoCs can be unavoidable when the system you're building on encourages design like that (.NET comes to mind, wouldn't be surprised if Android was like that as well), and does not necessarily mean that it's badly engineered.

createView... just awesome.

Well, writing a createView method itself isn't an issue unless you do it as a generic behaviour and for everything like in this Activity.

As far as I can see, that's lots of code for generating a UI. Interesting that they do that instead of using the usual xml files for specifying the UI on android; the result is that there's lots of code, a number of statements for each single UI element, and all the listeners and such are declared inline.

The compiled code will be fine, probably. It won't be much different from that would happen if they used the standard method of xml files and separate classes for that stuff, and might even be slightly faster.

But it surely is less maintainable, and the code definitely doesn't look generated to me. Maybe if you count copy-paste as code generation.

I hope they don't use Atom as an editor, because it surely won't be able to cope with such a file. [/s; but might even be true. not tested.]

Less maintainable? That's a nightmare. As an Android developer who has struggled with couple of 4 digit lines long activity classes I literally shuddered.

This is as if someone used a tool to translate UI code from some other platform to Android (completely into Java) and forgot to refactor.

Telegram has a poor security record, with homegrown encryption algorithms and no end to end encryption by default.

While I am wary of "build-your-own-crypto", saying it has a poor security record is, to my knowledge so far, incorrect. I have yet to see satisfactory cryptanalysis that shows a weakness (theoretical or real) in the protocol. If anything, it seems to withstand scrutiny so far. Even the dedicated security.se.com question[0] seems like it wraps itself in tautologies.

Until then, throwing the "homegrown encryption" argument to argue for Telegram's weakness is hand-waving and a straw man.

Telegram's security is sufficient for the threat model I have to face against now and in the future.

[0]: https://security.stackexchange.com/questions/49782/is-telegr...

FWIW, four years ago, the HN zeitgeist considered the cryptography blunders made by Telegram, when conbined with their hubris over how their protocol was clearly perfect, a death knell for the company from which they would never recover.

https://news.ycombinator.com/item?id=6948742

Indeed, I've been attacked and modded to oblivion on here for mentioning that I use Telegram, and the scope wasn't even security related, just an alternative to email for sending non-sensitive photos.

I've found Telegram to be more than good enough for day to day communications where I would previously have used email or SMS/MMS.

So is Signal and at least that is secure.

And once again, if I need that kind of security I'll seek it out. Telegram is more secure than email and SMS, and more convenient for my use case.

Signal is not feature rich and harder to get people to use. It also lacks the desktop/laptop integration.

Signal has desktop apps now https://signal.org/download/

How do they compare to the telegram desktop client? Are they electron apps?

Telegram is exactly as secure as email. SSL where third party email servers can see your traffic.

Signal is missing too many convenient features though, e.g. editing messages.

Why should one be able to edit a message after it is sent in a text-chat app?

Because one is human and one mistypes...

I'm normally on the Signal side of this argument, Telegram is trash and should not be used, but basic features like message editing are legitimate requests. I use Signal heavily and a number of times a day at least I or someone I'm chatting with corrects a previous poorly phrased or incorrectly typed messages. Because Signal has no mechanism for that, it usually involves retyping most of the message or the relevant word and adding an asterisk. I believe if Signal had built-in message editing support it would be used far more than I currently see corrections.

To be clear, I would expect Signal to allow me to see previous versions of a message and clearly indicate that it's been changed, but message editing is a valid feature that Signal should absolutely implement.

I've been retyping the offending word and adding * since I started using chat programs 20 years ago, and in every program since...I would think allowing messages to be edited after the fact would allow more problems than solving simple grammar or spelling mistakes most people can think around anyways.

I thought so, too. But after using Telegram for 3 years as my main messenger: No, it really is very convenient.

I'd imagine I would have thought the same thing before I actively used a messaging tool that supports message editing (I use Slack for work)

Further, the goal of Signal is not to work well for people who are accustomed to the various quirks of text-based communications that have existed since forever, but to provide a secure, modern communication client that normies don't find difficult to use. Message editing has become a standard feature of any usable modern messaging client.

I don't think I ever used a message service that allows you to edit messages.

Because people make (sometimes embarrassing) mistakes all the time. These circumstances can be saved if recipient hasn't read the message, and if they already have, can have a face-saving effect.

Embrace the mistakes and have a laugh!

And a desktop app that isn't Electron trash.

Signal requires to provide a phone number (read: disclose your identity to US Government) to use it.

I'm not a US citizen. And there is such a thing as prepaid phones, at least where I'm living.

Well, to me the whole thing is about a couple of things:

When Telegram first launched, people were reading their crypto whitepaper and going "Whoa, this is weird. You should probably not be doing it like this", and the reply was "Well, our 6 world champion coders (did you win a coding world championship?) think it is nice. Deal with it".

They then launched a bullshit crypto challenge (which would have been secure even using crypto primitives we _know_ are insecure). People told them that wasn't how it was done. They replied something cocky about world champion coders.

A couple of months later, someone found an gaping hole where the server could MITM every newly started secret chat (which their hack for forward secrecy a couple of years later would have made possible for every 100 messages).

I think their attitude towards encrypted messaging hasn't left puberty yet, and I recommend against it for everyone looking for a secure messenger. For anyone looking for a more convenient whatsapp without caring much for privacy by default, I don't mind recommending Telegram.

Regarding your future threat model: I certainly can't say whether having a copy of a significant chunk of my private communication on a server in Russia is a future threat to me. On the other hand, I don't use any social media, and have proper data retention policies on most of my online communications.

>having a copy of a significant chunk of my private communication on a server in Russia

To be fair, Telegram doesn't have servers in Russia.

"I have yet to see satisfactory cryptanalysis that shows a weakness (theoretical or real) in the protocol":

https://eprint.iacr.org/2015/1177.pdf

Can you mention something more specific, than suspicion to "homegrown encryption algorithms" that is just a yet another combination of DH-RSA-AES ?

https://eprint.iacr.org/2015/1177.pdf

And by the way, "no end to end encryption by default" -- is intentional tradeoff for cross-device message history sharing by default, which is not possible in case of e2e, without key transferring

I use a messaging service called Wire that does E2E, cross device messaging history by default (up to 6 devices), using very friendly, familiar UX - you only need to sign in on your device.

It might have been an intentional tradeoff when Telegram came out, but it's not any more. It's not a great defense for the fact that you chat history exists in plain text on Telegram servers.

Well for me, I prefer the way telegram has decided to handle end to end encryption. Let's face it - me and my friends have no use for it, and never will. Our threat model just doesn't really care about actors like the NSA hacking us. If they wanted there's tons of other ways to get our information any (like Google or Facebook accounts). But not being encrypted gives us features not possible otherwise, like url prefetching from the server or (when telegram was released, I guess it can be done now) accessing the chat from as many devices as you want. Yes, there is a tradeoff, but really I don't think most users are worried, nor should they be.

Speak for yourself. I care about privacy, and I couldn't give a shit about something so minor as link prefetching. I suspect I'm far from alone.

Anon1096 clearly stated that he was speaking for himself (or herself) but I agree with him: I want privacy but I don't necessarily need complete secrecy. This isn't a zero sum game: Telegram and Signal can coexist and serve different needs.

"I really don't think most users are worried" is the stark OPPOSITE of speaking for yourself.

While I do find such features convenient, more importantly to me they draw people to a more secure and privacy conscious messaging platform than the ultra-popular alternatives all the while allowing to upgrade to E2E† on demand. Signal completely fails in that regard, as the number of contacts (actual or potential) on it is precisely zero for me.

And you're very welcome to use something that fits your threat model.

I really don't understand the "hate" for Telegram on HN. I understand that its not (allegedly) as secure as signal for example, but like you said, state actors etc have other ways of getting that info (see xkcd [1]).

I personally use Telegram when talking to my wife because it lets me have a client on every platform i use and the messages follow me. I could do that with some of the other platforms as well, but im already on this platform, the wife is on the platform, and some other family members as well. Now would it be better if it was "more" secure? Of course it could, I mean, we could all be communicating with PGP as well, but most people don't because its a pain to use. So to me personally, Telegram is a useful middle ground between great security and usability.

As a added bonus, using the bot framework that Telegram offers, i have started creating a home automation bot. The framework is really fun and easy to use.

[1] https://xkcd.com/538/

> So to me personally, Telegram is a useful middle ground between great security and usability.

Same. I tried Signal a couple of years ago but the iOS app was prone to crashing, corrupting messages, and just plain losing words. Not really what you want.

> The framework is really fun and easy to use.

Yeah, I've done a couple of bots for it and it's pretty handy.

Telegram for me is the right trade off between security and convenience.

Ok. Lets see

https://medium.com/@wireapp/making-your-conversations-secure... does not state any information on key management solution: how encryption key is transferred to another device.

But it has links to https://en.wikipedia.org/wiki/Double_Ratchet_Algorithm which has nothing about cross device key sharing and https://wire.com/resource/Wire%20Security%20Whitepaper/downl.... I've read it through and found nothing on cross-device key sharing as well. Can you point me to the right part?

I'm flattered you assume I have this knowledge, but I don't. I would presume the key is derived from your sign-in information, but I honestly don't know.

What I do know is that you don't have access to messages that happened before you first signed in to the device you're using. My message history is complete on the phone I created my Wire account on, but I installed it on a laptop a few days later and as such is missing those first days. Messages get delivered to all your active devices without problem though.

The source code for all their different client varieties (iOS, Android, Web/Electron) and server software can be found here if you're interested [0].

[0]: https://github.com/wireapp

> the fact that you chat history exists in plain text on Telegram servers.

Source? https://telegram.org/blog/sessions-and-2-step-verification says:

"It allows you to set up a password that will be required every time you log into your account from a new device – in addition to the code you get in the SMS.

Be careful though: if you forget this password, you won't be able to access your messages from other devices."

This sounds to me that the chat history is encrypted with the password which doesn't leave my devices and therefore I'm not able to recover the history if I lose it. Since all clients are open-source, it should be possible to verify this.

edit: Okay the option for a recovery email could mean that they still have the password (or a key derived of it) on their servers - so basically plain-text.

> cross-device message history sharing by default, which is not possible in case of e2e, without key transferring

iMessage does that† (is that what you mean by "key transferring"?). I wish it could display the device key fingerprints somewhere, allow for key pinning, as well as offline key exchange between parties though††.

† in fact history is not centrally shared: messages are encrypted and published once per receiving device by the sender.

†† And be cross-platform. But the optimist in me says it's to control the security experience while the cynical in me says it's about network effects (although mitigated by SMS integration) for commercial reasons.

You're just repeating something that you read on HN. Entirely false.

False? Is it? I've yet to see a satisfactory answer to the criticisms that were posed about the security of Telegram.

Not entirely false, no. It does not have end-to-end encryption by default, and not at all for group chats, and the cryptography is a bigger unknown than with others.

> It does not have end-to-end encryption by default

That doesn't mean it "has a poor security record" (implying that the security is broken), it means it is not as secure as other options.

> not at all for group chats

May not be a requirement for group chats

> the cryptography is a bigger unknown than with others

Unknown does not mean bad, it means unknown. Almost definitely worse than Signal, but almost definitely better than Facebook Messenger.

To say Telegram is less secure than e.g. Signal is true. To say it "has a poor security record" is disingenuous and misleading. Once Telegram has had several security breaches, then that would be a fair phrase. Until then, there are levels of security for given threat models, and Telegram's is not as secure as others', by design.

Quoting Moxie: THere's nothing to be broken because there's no end-to-end encryption to begin with. Do you think nation state hackers publish their finding about the trove of private messages telegram's vast server complex holds indefinitely? hahahahaha

Are they allowing you to have accounts that don't depend on your phone number yet? I think last I checked you could only login with your phone number, which I consider extremely insecure, especially against a nation state agent (you'd think the Telegram team would care more about that, considering where they come from - and in fact, I believe there were some reports of Russian government hacking people like this with the help of national carriers, too).

> (you'd think the Telegram team would care more about that, considering where they come from)

That assumes the Telegram team aren't part of the nation state apparatus, surely.

Do you have any reason to assume they are? Of course, other than the fact that the founder is _RUSSIAN_ gasp

I don't assume they are. But equally I don't automatically assume they aren't, either.

You're just evading my point. Implicitely or explicitely you must have a _working_ hypothesis. When you make decisions, you must assume one or the other.

There is an optional two-factor login (phone+password).

Very few people use Telegram because of its security. Literally nobody of my contacts even uses the secret chat feature. They didn’t care when WhatsApp introduced e2e either.

People(IME at least) use Telegram(or WhatsApp) because it’s a nice to use app, both mobile and desktop.

Yes, the best thing about Telegram is that it doesn’t treat the desktop as an afterthought. Its desktop clients are first class true native apps, not be usual hot mess of half baked web/electron things that are offered to desktop users these days.

The iOS version of Telegram and Telegram X are currently not available in the Appstore.

It should be resolved soon: https://twitter.com/durov/status/958990254396059648

There should be a notification "your app is about to be removed, please contact us to solve the issue." (I'm assuming there wasn't because of other horror stories I've read.) If my business's service was suddenly down for 2-3 days due to some externality I couldn't immediately correct, it would be a death knell.

This is such bullshit, I would like see apple treat whatsapp like this.

Telegram is such an awesome software, I feel it's such a shame it's not more popular.

> P.S. Today, we're also introducing Telegram X for Android, a new official app.

What is wrong with the current app? Or was that not official? I find it better than the Facebook Messenger (light or not) and it's definitely much better than the Hangouts app from Google.

In the linked article in the quote it sounds like it's basically the same concept as Firefox Nightly. New and experimental features that may or may not make it into the other app. It's not meant to replace the old app right now, but it might do so in the future.

As they state that's because of lack of innovations among competitors. So they now compete internally.

They have had two apps (Telegram vs Telegram Desktop) on macOS too for a while. IIRC one is Electron-ish-based and the other is native Cocoa. The first one got features slightly faster but lacked Secret Chats IIRC.

I was actually positively surprised to see the Telegram application is qt based. Not sure which one is it of the two you mention, it's just called "Telegram", and is same on both Windows and OSX.

Explains why it doesn't take a minute to start and doesn't need a gigabyte of memory.

Also, I was very surprised to see that it was really QT based in the sense that there was no glue to an attached Chromium or something, just pure QT in C++.

The effort that went into that thing is amazing. It's fast on my 5 years old Pentium.

Which one should I be using? MacOS specific one feels polished whereas the Telegram for Desktop one doesn't have this strange issue where media sometimes never sends.

I prefer the macOS native client.

https://news.ycombinator.com/item?id=16280360

There is ongoing litigation with regards to the ownership of original Telegram client (issued under "Telegram LLC" based in Russia). Telegram X is released under "Telgram Messenger LLP" which is based in London and seemingly wholly owned by Durov. See my root comment for more details

Title is misleading, I think. You can only build your own client.

Reply on news.ycombinator.com