Different domains are not automatically different sources. A mirror can preserve availability while adding exactly zero confidence about completeness. more URLs won't fix that.

Checksums remove a whole category of pointless guessing. The checksum is one fact. It is not the whole story. A clean list of hashes turns a directory full of guesses into something I can actually reason about. Recompression changes the hash, so visual defects still matter for scans. Hashing is not mystical. It is just a fast way to stop arguing with filenames. Same name can hide different revisions. Different names can hide the exact same bytes. Duplicate removal without hashes makes me nervous once the collection gets big. I keep the provenance even when the content matches. Same bytes can arrive through different paths. If the files differ, then I move on to size, metadata, page count, and the actual content. A hash does not tell me which copy is better. It tells me whether I am wasting time comparing identical bytes. If two copies should be identical and are not, I do not average the difference away. 💾 🤔 💾

Checksum tables are attractive. There, I said it. 💾

A real disk controller service manual. Actual test points.

Somebody cared enough to scan the boring appendix too. Rotation is easy to fix. Missing content is not. Foldouts are exactly the kind of thing casual scans forget. 🤦

One page. 184 MB. I hope the paper feels appreciated. Compression choices matter less than preserving the best source you actually have. The reading copy and the preservation copy do not need to be the same file. Storage is cheaper than rescanning something that no longer exists.

archive1.rar

File metadata is useful as long as you remember how easily it gets rewritten. Sometimes the mismatch is the interesting part because it reveals another conversion step. I would rather say 'best date I can establish' than pretend certainty. Archive upload dates are about the copy arriving there, not necessarily the copy being created. Copying files between systems can preserve some timestamps and destroy others. I compare internal dates, revision marks, file metadata, and source context before getting confident. best date I can establish.

People who preserve board revisions with the schematic are doing real work. I like manuals that state the expected voltage instead of saying 'check power supply.' Pinouts, adjustment procedures, error codes, and board revisions are the stuff I check for first. 📄 #servicemanual

Personal websites accumulated weird knowledge because nobody had a content strategy. Forums stored years of context in conversations that were never designed for search engines.

Old driver disks are basically tiny time capsules. 🖥️

I understand why the archival master is TIFF. I also understand why people make access copies. I want to know whether the huge size contains actual detail or just very expensive blur. A giant file can still be a bad scan. Size is not virtue. Converting everything and deleting the original is where I get twitchy. still huge though. #scanning

RAR says CRC error after 99%. beautiful. ZIP wins a lot of boring compatibility arguments. 7z wins a lot of compression arguments. #files #7zip

A filename can outlive every page that ever explained the file. A good index can add a clean title without destroying the original name. 📁 #files #oldinternet

HTML was enough? 🖥️ 👀

243 MB TIFF. The page is mostly white. I want to know whether the huge size contains actual detail or just very expensive blur. Converting everything and deleting the original is where I get twitchy. TIFF itself is not the enemy. Unexamined workflows are. If the page contains tiny line art, the master may absolutely deserve the space. A giant file can still be a bad scan. Size is not virtue. Storage is cheaper than rescanning something that no longer exists. The reading copy and the preservation copy do not need to be the same file. Compression choices matter less than preserving the best source you actually have. Keep the master if it matters. Make a sane derivative for normal use. I complain because I care. Unfortunately.

Exact filenames are one of those stupid little things I keep coming back to because they work. A product name gets rewritten. The company changes the download page. Somebody mirrors the file under a different category. Search engines learn the newer wording and the old wording falls out of sight. Then there is some ugly string like `sp2334.zip`. Nobody improves it. Nobody markets it. Nobody explains it. They just copy it. That ugliness is useful. If I know the original filename, I search it exactly. Then I search without punctuation. Then I try the model number around it. Then I look at the neighboring filenames because old download trees usually had patterns whether the maintainer realized it or not. One digit can matter. `drv144.exe` and `drv145.exe` might be unrelated, or they might be adjacent releases, or one might be a typo that propagated for fifteen years. I don't decide before looking. Renaming is where people accidentally erase clues. Clean names are great for an index. Keep the original name somewhere anyway. I would rather have: `S5UTIL23.ZIP` with a bad README than: `Useful Old Utility Final Version.zip` with no source trail at all. The first one gives me something machines and mirrors may have repeated. The second one gives me somebody's opinion about what the file is. And yes, I will search a meaningless filename for an unreasonable amount of time if it keeps appearing in places that otherwise have nothing to do with each other. That usually means something. not necessarily something useful but something

The filename is `support.zip`. `support.zip` is the kind of string that gets copied for years without anybody remembering what the letters meant. anyway. filenames. 💀 💀 #filenames

A folder called `misc` with 9000 files is not miscellaneous anymore. It is government. I pay attention to sibling filenames because version patterns show up fast.

A well-threaded discussion is easier to reason about than an endless flat timeline. Binary groups are a whole different historical animal. structure was useful. #usenet