Long term strategies to address bandwidth issues and running costs: MangaDex@Home + Affiliates + BTC

[ixlone]

@TheDragonLord

Someone with more technical knowledge will need to answer that.

Current archive size is around 10TB.


@ripdog

The RC8+ clients we released today have fixed a lot of the issues and we are now hammering through requests and data-saver support is added.


@KazuyaKan

Current IP will be fine.

 

[blackyawgdom]

@Seneca7C
Theyve already said those requirements are just for the initial rollout. They wanted to focus on high quality/high speed people first because there's only so many security certificates they can assign per day/hour/whatever the limit is. Once everything reaches Stage 2 (Meaning they aren't pushing a bugfix every 3 hours) the bandwith requirements will be lowered

 

[knabber]

Is there a way to update a client request? or should I just request a new client?
Set the bandwidth too low in the initial request.

 

[ilx]

Thanks @BraveDude8 and @Holo for the info.

As the project isn't open for contributions yet, I'd like to point out a few things regarding malicious nodes for the issue tracker on the forum before it's rolled out in production:
- A node could reply with HTTP 401 and a HTTP Basic Authentication message which will cause a pop-up for the user with some text that the attacker controls. This can be fixed by setting the crossorigin attribute to anonymous and setting the Access-Control-Allow-Origin header in the HTTP response of the nodes.
- The served image contents can be tampered with. We can verify the contents by providing the visitor with the correct hash and comparing it to the hashed contents of the received image in the browser with JS. Unfortunately, we can't make use of subresource integrity because the only supported types for that tag is js and css, so we have to do it manually in JS. The implementation of this most likely involves creating an img DOM element that refers to an authentic byte array in the form of a JavaScript Blob object. Implementing it in this way would also fix the HTTP Basic Authentication issue mentioned earlier in addition to asserting the integrity of the image.

 

[bobjrsenior]

@creatix thanks for letting me know. It's probably for the best that it's encrypted.

 

[Aereus]

@XXXXXXXXXIII You can already do that now. There are several utilities out there that support scraping chapters from Mangadex.

 

[knabber]

@ilx
Probably would be simpler to just report a image or page load that is causing issues? Reports against a node go over a threshold, node is removed from the active pool.
This would also work for nodes that fail to load image or otherwise misbehaving. As far as I remember HatH does this, lots of users hitting the "reload broken image" will cause your node's trust to tank.

As far as I am aware, the issues that you describe have not been an issue in the HatH network for the last 8+ years.

 

[gagahpangeran]

I have small digital ocean droplet with Singapore server and only run small nodejs project in there. The upload and download speed is about 900 Mbps (testing using speedtest cli) but the storage is only 25GB and now it's 40% used. Looks like I can't help for now (or can I?).
I live in SEA country and my upload and download speed is about 15 Mbps. Mostly I used this server as a proxy to Mangadex and the result is very very fast load time compare to without using my server as a proxy.

 

[Holo]

Quick update for tonight - the network is coping very well now. Every user and guest are now using MD@H (unless you switched off the option).

 

[Lancet11]

So just curious, for the funding portion of Mangadex, would it be possible to set up a Paypal account that's not affiliated directly with MD then the person(somebody from MD) running that account can just move the funds over to crypto then towards MD? More steps but once it hits Crypto they really can't trace it thus allowing PayPal donations again

 

[blackyawgdom]

@Lancet11 That would be flagged really really fast as money laundering

Paypal (absolute worst) AI.
This account has hundreds of small transfers into it and its all heading to a crypto exchange.... *pings the FBI/CIA/IRS*

 

[wedge1001]

@Lancet11
what @blackyawgdom says.

What's possible is that someone (perhaps i will do this) provides some services
e.g. adding another M@/h server and you can buy a slice of it.
This can then be payed with paypal.

Would someone be interested in such a deal?

 

[RonBWL]

@wedge1001 Pretty sure there was a solution for server donations, in which individual supporters would rent one and then let MD access it. Technically speaking said individual could ask for PP donations and as long as readers know they donate to the right person, it could work.

 

[Samhill]

@ixlone @Holo

MAH is serving up some images from a previous chapter in the next chapter.

 

[Holo]

Can you link the chapters involved so we can investigate?

 

[Samhill]

@Holo

https://mangadex.org/chapter/820176/1
https://mangadex.org/chapter/877350/1

I used long strip. It also happened once before in that series, six and seven.
https://mangadex.org/chapter/816890/1
https://mangadex.org/chapter/816942/1

 

[kolorafa]

@samhill
I had similar thing happen (maybe the same happen to you), it was due to switching to next chapter before all the loading was finished, so the chapter did change but the image from previous was loaded after the change and img was put in currently open chapter.

It would be nice to include check after img load if the chapter didn't change while loading.

 

[Holo]

@teasday thoughts?

 

[Afiaki]

Regarding the bug about [images loading in the next chapter because they haven't finished loading], that's a reader issue which has been happening for a very long time before MDH.

 

[kolorafa]

Browsers have a tendency to limit and queue simultaneous connection into same host and/or ip, as MDH now have different ones it might trigger the error more often as MD Api and MDH images will be handled in different queue.

And I think the reader bug could be fixed by testing if the chapter match before replacing img src after img load, it shouldn't be hard to do, and (at least me) would really appreciate that fix

(And yes, that bug was way before MDH)