New Filebase (introduced in v3.19)

The new filebase files are stored in the same database location as before (e.g. data/dirs/), but the file extensions and storage formats are different:

Purpose Old New
Index (e.g. filenames) code.ixb code.sid
Data (e.g. descriptions) code.dat code.shd
Extended Descriptions code.exb code.sdt
Metadata code.dab code.ini
Allocation Tables N/A code.sda and code.sha

If these new filebase file extensions look familiar to you, that's because we're using the Synchronet Message Base format/library (v3.0 now) for the underlying database. This means that the SMB tools you may be familiar with (e.g. CHKSMB, FIXSMB, SMBUTIL) also work on the new filebases.

The conversion of the filebases to the new format should occur automatically when you run 'jsexec update' which in turn will execute the program 'upgrade_to_v319' when appropriate (just one time). Once converted, you can delete the old files or leave them in place in case you need to revert back to Synchronet v3.18 for any reason. The old filebase files won't harm anything if left.

The creation of each new filebase will automatically calculate and store the hashes of the contents of the actual files available for download. These hashes are useful for duplicate file detection and data integrity assurance. If you wish to opt-out of the file hashing (which consumes the majority of the time during the upgrade process), you can turn off file hashing in the per-directory Toggle Options in SCFG->File Areas. You would have to perform this opt-out for the directories of choice before you run 'jsexec update' / 'upgrade_to_v319'. You should not normally need to run 'upgrade_to_v319'.

Long Filenames

While filenames stored in the filebases used to be limited to MS-DOS compatible 8.3 formatted names, longer filenames are now supported on all platforms. Additionally, some previously invalid filename characters (e.g. spaces) are now allowed and files without extensions (i.e. no '.' in their filename are now supported).

Although Synchronet for Windows previously used Win32 API functions for short ↔ long filename conversions in the Windows builds of Synchronet, resulting in the unfortunate Micros~1 shorted filenames stored and sometimes seen, that is no longer the case. Except for the %~ command-line specifier, those short/long filename conversion functions are no longer in use anywhere within Synchronet for Windows - it's native filenames through-out. The filebase conversion process (upgrade_to_v319 / 'jsexec update') on Windows will attempt to automatically resolve the native/long filenames and store those names and only those names in the new filebases.

Note: abbreviated versions of long filenames are displayed in some situations to accommodate the limited width of a traditional BBS user terminal. An effort is made to always display the full file extension/suffix however (e.g. “longfilename.jpeg” may be displayed as “longfil.jpeg”).

Note: only 64 characters of each filename (always including any extension) are indexed for searches and duplicate checking, but the entire filename, up to 64K characters in length, is stored intact in the filebase.

Filenames with extensions longer than 3 characters, e.g. “.jpeg”, “.tar.gz”, can be added to the filebases, but the configurable compressible, extractable, and viewable file types/extensions remain limited to 3 characters in SCFG. Similarly, a maximum length of 3 character archive “types” are stored per BBS user record (for each user's QWK packet format and temp archive preference).

Since support for the viewing and extraction of several common archive formats is now built into Synchronet, files uploaded with unrecognized or unsupported file extensions (e.g. “.tar.gz” instead of “.tgz”) may still be successfully extracted or viewed.

Large Files

Files greater in size than 2GB or 4GB (depending) were previously a problem. Though there are still some 32-bit file length limitations (e.g. only files smaller than 4GB in size will be hashed), there is better and increasing support for larger files in general.

Note: the ZMODEM transfer protocol as designed by Chuck Forsberg only supports files up to 4GB in size and in many cases, files greater than or equal to 2GB in size will prove difficult or impossible to transfer between some ZMODEM implementations. In general, it is recommended to use an alternate transfer protocol (e.g. YMODEM[-G], FTP, HTTP) for files >= 2GB in length.

Large Directories

Due to the new filebase design, directories with more than 10,000 files are now supported (though, not encouraged).

Descriptions

The file “summary” or single-line “short description” remains limited to 58 characters in length for practical purposes. Though longer file summaries (up to 64KB) can be stored in the filebase, they are not recommended.

Extended (multi-line) descriptions may now span more than the previous limit of 10 lines or 512 total characters. There is no technical limit to the length of extended file descriptions, though a limit of 4000 characters imported from description files embedded in archives (e.g. FILE_ID.ANS, FILE_ID.DIZ, or DESC.SDI) is being imposed.

Batches

File upload and download batch queues used to be maintained in memory (though they were written to disk files to be retained between user logons), they are now entirely maintained in disk files (data/user/*.upload and *.dnload in .ini file format). This means that custom batch management can now be performed easily by modules or non-Terminal Server scripts.

Hashes

Files are now hashed, by default, using multiple hashing algorithms (CRC16, CRC32, MD5, and SHA1) for duplicate file detection and for reporting to users (e.g. to insure data integrity). For a file to be considered a duplicate (i.e. and rejected for upload) it must have the same size and hash values as another file in a filebase already. Each directory is configurable as to whether or not to hash its files or use it for duplicate file detection (by name or hash).

Sorting

While the old filebase indexes were re-ordered/sorted whenever a new file was added, the new filebase indexes are sorted dynamically, when appropriate (e.g. when viewed).

As a result, the “RESORT” file transfer operator command has been removed.

The new filebases are naturally indexed in the order in which the files are imported into the database. Sorting of the files for display purposes in the terminal and FTP servers is optional and configured by the sysop:

Name Ascending (case-insensitive)
Name Descending (case-insensitive)
Name Ascending (case-sensitive)
Name Descending (case-sensitive)
Date Ascending
Date Descending

Tags

Individual files can now be tagged (with tag-words, ala hashtags) for easy searching/grouping. This feature will be utilized/enhanced more in the future.

JavaScript

The new “FileBase” class is used (similar to the existing MsgBase class) to open and access filebases from JavaScript modules. Using this class, the default methods of listing and transferring files can be replaced with custom modules.

TickIT

The new FileBase JS class is now used to import files directly from FidoNet -style .TIC files (via tickit.js) so no dependency or invocation of any external utilities (e.g. addfiles) is required.

Utilities

The native utilities ADDFILES, FILELIST, DELFILES, and DUPEFIND have been replaced with similarly named and purposed JavaScript utility scripts to be invoked with JSexec or via timed event:

  • addfiles.js for importing lists of files into filebases
  • postfile.js for importing a single file into a filebase
  • filelist.js for generating file listings from filebases
  • delfiles.js for removing files from filebases
  • dupefind.js for detecting/listing/removing duplicate files in the filebases

Performance

Due to the nature and use of the new filebase API, file listings are much faster (e.g. large file listings from the Synchronet FTP server) as well as improved performance when filename/pattern, description text, and duplicate file searching.

FTP Server

Due to the removal of support for rendering FTP-downloaded content (e.g. HTML files) in modern web browsers, the FTP Server no longer supports dynamic HTML index file generation (e.g. 00index.html). Instead, we will focus on better support for filebase browsing and file transfers via HTTP and HTTPS in addition to the traditional FTP and FTPS uses. The dynamic generation of the ASCII file listings via FTP (e.g. 00index) is still supported by the FTP server, though now much faster than before.

libarchive

The libarchive library has now been integrated into Synchronet (and exposed via the new “Archive” JavaScript class) and integrated into SBBSecho so that the creation, listing/viewing, and extraction of archived files can now be performed “in-process” without the invocation of or dependency on external programs (e.g. Info-Zip unzip or PKUNZIP).

Formats fully supported:

  • zip
  • 7zip
  • gzipped-tar
  • bzipped-tar

Formats supported for viewing and extraction only:

  • rar (original and RAR v5 format)
  • lha/lzh
  • iso
  • xar
  • cab

This means that for most BBSes, no “Compressible” or “Extractable” file types need to be configured in SCFG->File Options. Additionally, by setting “Archive Format” to “ZIP” for SCFG->Networks->QWK->Hubs, no “pack” or “unpack” command-line need be configured.

For listing the contents of archives, the new exec/archive.js utility script may be installed as a “Viewable File Type” handler for the commonly supported file extensions by running 'jsexec archive.js install'. This installs archive.js as a final “fall-through” viewer for “*” files (all file types).

DIZ

Description files embedded in archives (e.g. FILE_ID.DIZ) are now supported more uniformly and seamlessly. Supported ANSI sequences (e.g. attribute/color changes) are now converted to Ctrl-A equivalents upon import. Metadata (e.g. declaring the file's author or group) is now imported from SAUCE records of DIZ files, when it exists.

File Echoes

Each file transfer directory configured in SCFG->File Areas may now have an “Area Tag” explicitly set for FidoNet-style file distribution networks. If an Area Tag is not explicitly set, then the directory's short name is used (with spaces replaced with underscores) automatically. tickit.js now uses this new “area_tag” file_area.dir[] JS property for its “AutoAreas” feature.

Opened Files

The “open” (reference) counter for files is now gone. If you want to remove a file from the filebase while a user has it in their batch download queue or is actively downloading it, nothing is preventing you from doing so.

As a result, the “CLOSE” file transfer operator command has been removed.

Alternate File Paths

The support for “Alternate File Paths” has been removed. There are better modern operating/file system solutions to the original problem solved with this feature.

As a result, the “ALTUL” file transfer operator command has been removed.

Bi-directional File Transfers

The protocol drivers that supported bi-directional file transfers (Bi-Modem, HS/Link) are now long unsupported DOS/OS2 programs with no equivalent in the modern world. Bye bye Bi-modem. :-(

User-to-user File Transfers

While user-to-user file transfers are supported in the new file base, the “to” users of existing user-to-user transfers (in the “user” file area, if you have one) will not be migrated by 'upgrade_to_v319' to the new filebase.

A quick way to check if this will impact you is to see if you have a non-zero-length data/xfer.ixt file. If you do, then you have some pending user-to-user transfers that you may want to deal with.

See Also

history/newfilebase.txt · Last modified: 2023/02/28 14:54 by digital man
Back to top
CC Attribution 4.0 International
Driven by DokuWiki Recent changes RSS feed Valid CSS Valid XHTML 1.0