AstroSkipper

Thanks @nicolaasjan! I know that. We both had a conversation with maroc in those days. At that time, I have installed MPC-HC 2.2.1. But I prefer the combination I mentioned in my previous post. And I enhanced ytBATCH for Windows XP to open my files downloaded from YouTube with ffplay. And for only watching YouTube streams, VLC with my youtube.lua script is absolutely fine.

After all the drama surrounding MSFN, Google has gone and made things even worse. YouTube no longer runs at all within browsers on older, low-performance hardware like my P4. The Vorapis script can no longer play videos on my hardware in New Moon 28 and Serpent 52 on Windows XP. For some reason, the h264 codec can no longer be used in the same way as before. So I’ve removed all scripts and extensions from these browsers. They’ve become useless to me. I now only use VLC with my youtube.lua script. I search on YouTube using the 3D YouTube Downloader, which can of course also download videos and audio files. And naturally, I’m using the latest version of my ‘ytBATCH for Windows XP’ (version 1.7).

For a longer time, New Noon 28 and Serpent 52 don't work correctly on MediaFire. When logging in and opening a folder with more files the screen can show there is no scroll bar to get the next ones in the list. Here are screenshots from the NM and Sp browser consoles: And here is a screenshot of a MediaFire's folder taken from my account where you can see the missing vertical scroll bar:

VLC is also used to watch streams. And that means YouTube streams, too. Therefore, your request is totally on topic. Your observation in terms of the 3.0.8 version is interesting. I will try this version in my system the next days. One thing is clear. The latest versions aren’t always the best on older systems. They are no longer tested on older operating systems. I can confirm that regarding your BBC Radio 2 Live stream http://as-hls-ww-live.akamaized.net/pool_74208725/live/ww/bbc_radio_two/bbc_radio_two.isml/bbc_radio_two-audio=96000.m3u8 the 3.0.24 version from 29.05.2026 behaves very bad, but the 3.0.8 version perfectly. Even the 3.0.20 version is ok but the 3.0.8 version performs best. When it comes to my YouTube test video https://www.youtube.com/watch?v=ffcitRgiNDs however, I can't see any differences in CPU usage.

VLC is also used to watch streams. And that means YouTube streams, too. Therefore, your request is totally on topic. Your observation in terms of the 3.0.8 version is interesting. I will try this version in my system the next days. One thing is clear. The latest versions aren’t always the best on older systems. They are no longer tested on older operating systems.

I realise VLC specifically is NOT the topic of this thread, but could you be so kind as to actually pinpoint/link to that version? On https://artifacts.videolan.org/vlc-3.0/nightly-win32/ you can see that their server retention goes back to last December, ONLY, so I expect, in the future, that older "artifacts" will be forever purged ; perhaps this last, XP-compatible, build deserves to be archived for posterity ... All 3.0.23 versions are compatible with Windows XP. The last 3.0.23 version was releases on 12.05.2026. The last Windows XP compatible 3.0.24 version was from 29.05.2026.

Yep, you're right. I thought he meant av1. Anyway, my youtube.luac is requesting the h264 codec. And my tests were done using yt-dlp from 2026.06.10, so not the recent one.

I have even tested my youtube.luac file in the last XP-compatible VLC beta 3.0.24 (the very last recent versions are no longer compatible with Windows XP 😟) and @nicolaasjan's yt-dlp version from 10.06.2026.

Today, I have tested my youtube.lua script again. And it is working in VLC v3.0.23 without any problems. No errors, no loops and no need to click the Play button more times. On older hardware, the video format h264 is the best option. That's why my script uses this codec. Here is my last code again with the bracket correction mentioned by @Ben Markson: -- YouTube Link Resolver for VLC with Separate Video and Audio URLs -- Place this script in VLC's lua/playlist directory local yt_dlp_path = 'yt-dlp.exe' local yt_dlp_silent_path = 'hidecon.exe' function sleep(s) local ntime = os.time() + s repeat until os.time() > ntime end function probe() -- Check if the input is a YouTube link return (vlc.access == "http" or vlc.access == "https") and (string.match(vlc.path, "youtube%.com") or string.match(vlc.path, "youtu%.be")) end function parse() -- Construct the full YouTube URL local youtube_url = vlc.access .. "://" .. vlc.path -- Extract "quality" query parameter if present local quality = youtube_url:match("[&?]quality=(%d+)[pP]?") youtube_url = youtube_url:gsub("[&?]quality=%d+[pP]?", ""):gsub("[&?]$", "") -- Remove trailing ? or & -- Get the preferred resolution preset by the user in VLC local prefres = vlc.var.inherit(nil, "preferred-resolution") if prefres < 0 then prefres = 2160 -- Best quality set to 2160p to overwrite VLC's native value of -1 end local allowed_qualities = { ["240"] = true, ["360"] = true, ["480"] = true, ["720"] = true, ["1080"] = true, ["2160"] = true } -- Default quality limited to the preferred resolution taken from VLC's settings local format_string = string.format("bestvideo[height<=%i]+bestaudio", prefres) if quality and allowed_qualities[quality] then format_string = string.format("bestvideo[height<=%i]+bestaudio", quality) vlc.msg.info("Using requested quality: " .. quality .. "p") else vlc.msg.info("No valid quality specified. Defaulting to best available.") end local cmd_hidden = 'hidecon &' -- Start cmd hidden and ... -- No better codec than h264 (an important switch for older hardware). This can of course be changed by the users according to their needs. local codec_limit = '-S "codec:h264"' local video_url = '' local audio_url = '' local yt_dlp_silent_exists = io.open(yt_dlp_silent_path, "r") ~= nil if not yt_dlp_silent_exists then vlc.msg.info(yt_dlp_silent_path .. " not found. Falling back to " .. yt_dlp_path) cmd_hidden = 'PowerShell.exe -windowstyle hidden cmd /c &' -- Start cmd hidden and ... local cmd = string.format( '%s "%s" %s -f \"%s\" -g "%s"', cmd_hidden, yt_dlp_path, codec_limit, format_string, youtube_url ) local handle = io.popen(cmd) video_url = handle:read("*l") audio_url = handle:read("*l") handle:close() else vlc.msg.info(yt_dlp_silent_path .. " found. Running program") local cmd = string.format( '%s "%s" %s -f \"%s\" -g "%s"', cmd_hidden, yt_dlp_path, codec_limit, format_string, youtube_url ) local handle = io.popen(cmd) video_url = handle:read("*l") audio_url = handle:read("*l") handle:close() end video_url = video_url and video_url:gsub("^%s+", ""):gsub("%s+$", "") or "" audio_url = audio_url and audio_url:gsub("^%s+", ""):gsub("%s+$", "") or "" vlc.msg.info("[YouTube Resolver] Original URL: " .. youtube_url) vlc.msg.info("[YouTube Resolver] Video URL: " .. video_url) if audio_url and audio_url ~= "" then return { { path = video_url, name = vlc.path .. " (Video)", options = { ":input-slave=" .. audio_url } } } else return { { path = video_url, name = vlc.path .. " (Video + Audio)" } } end end And today I reduced and optimised my code again and compiled it to a LUAC file. Here is my most recent version for old, weak computers (using h264): https://www.mediafire.com/file/olbadgp55gzn6aj/youtube.luac/file Cheers, AstroSkipper

Thanks for your reply and hints, @jumper! Yesteday, I checked all prefs related to memory management and performance. And there are a lot. After investigating them, I adjusted my settings, and now New Moon 28 behaves better than before. When starting the browser under same condition as before, the palemoon.exe process decreases its RAM usage from 184 MB to 139 MB. And after surfing some sites and closing their tabs, my automatic memory minimization in the background is now able to release more memory than before. But of course, that doesn’t solve the fundamental problem with browsers that run in single-process mode. But it’s a significant improvement nonetheless. My user.js is now totally optimised for my old P4 processor. Regarding your suggestion, I don't think setting browser.cache.memory to false is a good idea. My IDE hard disk is too slow for that. Although my RAM is SD-RAM, it is much faster.

Yesterday, I uploaded files to MediaFire, and the used RAM increased to over 600 MB. After closing the tab, only a very, very small amount of used RAM was freed up. @roytam1, any chance to natively improve that behaviour in New Moon 28?

Update notification! As already reported, the Root Certificates have been updated and are now from 29-04-2026. Here again a screenshot: After a long delay, I have just updated my self-created, offline Root Certificate Updaters in the section 11.2.4. Downloads related to Root Certificate Updates (in the first post of this thread). Cheers, AstroSkipper

The only real issue that all browsers in single-process mode as, for example, New Moon 28, have – and which has always been the case with every version of Firefox – is their inability to free up RAM when tabs are unloaded or closed. The amount of memory that is actually released is pitiful compared to the amount that was originally used. It’s a problem that has remained unresolved since the very beginning, so to speak. And these days, developers no longer worry about wasting RAM, because most of them have plenty of it. But my Windows XP computer doesn't. What is particularly pleasing, however, is that GitHub has been running smoothly in New Moon 28 for some time now and, as far as I can tell, is actually running very well at the moment.

Here's a quick status update on New Moon 28 (32-bit) (2026-06-04) - the current release - under Windows XP running on a single-core 32-bit P4 with 1.5 GB of RAM. I have installed 25 extensions in my profile, 23 of which are enabled, 17 UC.JS scripts, 1 UC.XUL script, 4 USER.JS scripts, and 4 CSS stylesheets. When I launch the browser, it uses 184 MB of RAM, and it takes only a few seconds to load. My browser console shows no errors when the browser starts. If error messages appear, they are caused by websites. The reason for this is that New Moon 28 runs smoothly out of the box – at least as far as the browser console is concerned, thanks to @roytam1 – and I wrote my UC.JS scripts on the premise of running completely error-free. And regarding extensions I either fixed faulty ones myself or, if unfixable, I removed them.

It's good to hear you're doing well. Have you managed to sort out the GitHub issue?

Ok. Thanks!

This is the closed, archived repository of the original uBlock Origin Legacy: https://github.com/gorhill/uBlock-for-firefox-legacy There, you can clearly see that there are no recent patches or updates for it. No more development. @UCyborg started to port patches from the webextension to the legacy version https://github.com/UCyborg/uBlock-for-firefox-legacy. But with long delays in development. His last version is nearly one year ago. And as I already reported, I noticed incompatibilities and higher resource usage when using his version. So, as long as my version is running without any issues, I won’t be making any changes or updates. And if I do, it will only be to the assets, i.e. the filter lists if there are issues. Be that as it may, @Monroe, @Amigafever and @Rod Steel, I'm still pleased to see that there is interest in my uBlick Origin Legacy build.

My main focus was on fixing all the obvious errors caused by a completely outdated XPI file and its incompatibility with the new format of the assets file. That is why I also changed the uBO ID in my version and created my own assets file. As I already said, it’s not easy to implement patches, and that also includes security patches, in legacy extensions that were actually developed for webextensions. So, if you are more interested in security patches of uBO itself, you should use @UCyborg's version and wait for the most recent ones if they are ever implemented. But if you are rather interested in compatibilty and better performance, you should use my version. So it’s entirely up to you.

Everything I reported earlier is already old news. The error has been back in full force for almost a week now. Views are no longer being counted at all. For example:

The @AstroSkipper version is still working. I created it back when nobody was maintaining the legacy version of uBO anymore. This had led to @UCyborg also working on an updated version again. I tested it for a while and actually think any further development is a good thing, but unfortunately I found incompatibilities and higher resource usage. That’s why I’ve gone back to my version. In my opinion, it’s not easy to implement features in legacy extensions that were actually developed for webextensions.

Yes, I used eTools.ch for a while too. But it produced the same rubbish as Google, as it seems to be using Google’s data. I’ve deleted it from my list of search engines.

I’m sorry to hear that, of course. I can well understand how you feel. To put it mildly, I’m not doing too well myself, personally and privately. Especially when it comes to health. I know exactly what you mean. I’m a one-man band in everything I do, too. To be honest, your personal life comes first. So the decision is entirely up to you. Thanks, but I'll decide that for myself. Have a nice night, too!

Since the administrator @xper apparently does not feel it necessary to officially thank his members for their active participation in providing data that led to a precise theory, on the basis of which it was possible to prove the infection of the MSFN server, I shall do so here. @EliraFriesnan, @roytam1, @NotHereToPlayGames, @Ben Markson, @modnar, @deomsh and @nicolaasjan. Thank you for your confirmations with search engine results you generated from all over the world! That was a real community effort at its very best. And since my ‘cloaking’ theory https://msfn.org/board/topic/187801-msfn-and-its-lack-of-search-results-in-google-bing-and-all-search-engines-depending-on-them/page/5/#findComment-1287780, which, although based on many facts, remains a theory until it has been subjected to rigorous, unequivocal and, above all, empirical verification, I would like to once again express my sincere thanks to @nicolaasjan for his clever experiment and for highlighting the malware infestation in this post: https://msfn.org/board/topic/187801-msfn-and-its-lack-of-search-results-in-google-bing-and-all-search-engines-depending-on-them/page/5/#findComment-1287787. This made the culprit clearly visible and made it impossible to continue ignoring or denying an infection of the MSFN server. In mathematics, this signifies the end of a proof, i.e. q.e.d. And that is the most important thing for me. Cleaning it up was then just a formality and, really, a piece of cake for someone with experience, if you already know exactly what you’re looking for and have been given all the necessary information. Unfortunately, it will now take many weeks for MSFN to recover in the Google index. Not to mention the disastrous drop in the rankings of all MSFN threads.

I don't mind a typo. I don't think @VistaLover owns a time machine either :

I don’t need to stroke my ego. I don’t need to do that, as I know exactly what I’m capable of. But it was simply a reminder of all the hard work I’ve put in since 21 May – work for which any expert would be paid, but which I’ve, of course, offered to the forum free of charge. Your comment is therefore inappropriate and unnecessary.