Rapid upload matrix

种子系统依赖网盘提供的哈希来避免重复下载。不同网盘「列表提供的哈希」与「秒传所需的哈希」各不相同。下面把这两个维度合并为一张二维矩阵：行 = 源驱动列表提供的哈希，列 = 目标驱动秒传所需的哈希，交叉格即「从源到目标能否秒传」。

The seed system relies on drive-provided hashes to avoid redundant downloads. Different drives differ in the hashes they "provide in listing" and "require for rapid upload". The two dimensions are merged below into a single two-dimensional matrix: rows = hashes provided by the source listing, columns = hashes required by the target for rapid upload; each cell answers "can this source rapid-upload to this target".

图例

Legend

符号	含义
✅	可直接秒传：源列表已提供目标所需的全部哈希，无需下载文件内容
🔶	需补算哈希：源提供了部分所需哈希，但缺少分片哈希或其它算法，需下载文件补算后才能秒传
❌	无法秒传：源缺少目标秒传所需的关键哈希
➖	不适用：目标驱动本身不支持秒传（只能普通上传）

Symbol	Meaning
✅	Direct rapid upload: the source listing already provides every hash the target needs, no content download
🔶	Needs hashing: the source provides part of the required hashes but is missing piece hashes or another algorithm, so the file must be downloaded to compute the rest first
❌	Cannot rapid upload: the source lacks the key hash the target requires
➖	Not applicable: the target drive does not support rapid upload (normal upload only)

秒传矩阵

Rapid upload matrix

源提供 \ 目标需要	MD5	MD5 + 分片	SHA1	MD5 + SHA1	SHA1 或 MD5	GCID	SHA256	无秒传
MD5	✅	🔶	❌	🔶	✅	❌	❌	➖
SHA1	❌	❌	✅	🔶	✅	❌	❌	➖
MD5 + SHA1	✅	🔶	✅	✅	✅	❌	❌	➖
MD5 + SHA1 + SHA256	✅	🔶	✅	✅	✅	❌	✅	➖
GCID（分块 SHA1）	❌	❌	❌	❌	❌	✅	❌	➖
无	❌	❌	❌	❌	❌	❌	❌	➖

Source provides \ Target needs	MD5	MD5 + pieces	SHA1	MD5 + SHA1	SHA1 or MD5	GCID	SHA256	no rapid upload
MD5	✅	🔶	❌	🔶	✅	❌	❌	➖
SHA1	❌	❌	✅	🔶	✅	❌	❌	➖
MD5 + SHA1	✅	🔶	✅	✅	✅	❌	❌	➖
MD5 + SHA1 + SHA256	✅	🔶	✅	✅	✅	❌	✅	➖
GCID (block SHA1)	❌	❌	❌	❌	❌	✅	❌	➖
none	❌	❌	❌	❌	❌	❌	❌	➖

驱动哈希能力映射

Drive hash capability mapping

下表把每个驱动映射到矩阵的行（列表提供）与列（秒传所需），据此即可在矩阵中定位任意「源 → 目标」组合。

驱动	列表提供哈希（源 / 行）	秒传所需哈希（目标 / 列）
天翼云盘 189pc	MD5	MD5 + 分片 MD5（`slice_md5`）
天翼云盘 189（旧版）	MD5	MD5 + 分片 MD5
天翼云盘 189tv	MD5	MD5
阿里云盘	SHA1	SHA1（`pre_hash` + `content_hash`）
阿里云盘 Open	SHA1	SHA1（`pre_hash`）
百度网盘	无（API 返回的 MD5 不可信）	MD5（`content-md5` + `slice-md5`）
PikPak	GCID（分块 SHA1）	GCID
115 网盘	SHA1	SHA1
115 Open	SHA1	SHA1
迅雷	GCID	GCID
迅雷 X	GCID	GCID
迅雷浏览器	GCID	GCID
夸克网盘 Open	SHA1	MD5 + SHA1
夸克网盘 UC	无	MD5 + SHA1
febbox	GCID	GCID
移动云盘 139	无（`digest` 未暴露）	SHA256（快传）
123 网盘	MD5（`etag`）	MD5
123 Open	SHA1 / MD5	SHA1（`sha1_reuse`）或 MD5
Google Drive	MD5 + SHA1 + SHA256	无秒传（普通上传）
OneDrive	无	无秒传（普通上传）

The table maps each drive to the matrix’s rows (listing provides) and columns (rapid upload needs), so any "source → target" pair can be located in the matrix.

Drive	Hashes provided (source / row)	Rapid upload requires (target / column)
189pc	MD5	MD5 + piece MD5 (`slice_md5`)
189 (legacy)	MD5	MD5 + piece MD5
189tv	MD5	MD5
Aliyundrive	SHA1	SHA1 (`pre_hash` + `content_hash`)
Aliyundrive Open	SHA1	SHA1 (`pre_hash`)
Baidu	none (returned MD5 is untrusted)	MD5 (`content-md5` + `slice-md5`)
PikPak	GCID (block SHA1)	GCID
115	SHA1	SHA1
115 Open	SHA1	SHA1
Thunder	GCID	GCID
Thunder X	GCID	GCID
Thunder Browser	GCID	GCID
Quark Open	SHA1	MD5 + SHA1
Quark UC	none	MD5 + SHA1
febbox	GCID	GCID
139 (China Mobile)	none (`digest` not exposed)	SHA256 (rapid upload)
123	MD5 (`etag`)	MD5
123 Open	SHA1 / MD5	SHA1 (`sha1_reuse`) or MD5
Google Drive	MD5 + SHA1 + SHA256	no rapid upload (normal upload)
OneDrive	none	no rapid upload (normal upload)

如何阅读矩阵

How to read the matrix

阿里云盘 → 115：源提供 SHA1，目标需 SHA1 → 矩阵 SHA1 × SHA1 = ✅ 可直接秒传。
123 网盘 → 天翼云 189pc：源提供 MD5，目标需 MD5 + 分片 → 矩阵 MD5 × MD5+分片 = 🔶 需下载补算分片 MD5。
PikPak → 阿里云盘：源提供 GCID，目标需 SHA1 → 矩阵 GCID × SHA1 = ❌ 无法秒传。
任意源 → Google Drive / OneDrive：目标无秒传 → 矩阵末列 = ➖，只能普通上传。

Aliyundrive → 115: source provides SHA1, target needs SHA1 → matrix SHA1 × SHA1 = ✅ direct rapid upload.
123 → 189pc: source provides MD5, target needs MD5 + pieces → matrix MD5 × MD5+pieces = 🔶 must download to compute piece MD5.
PikPak → Aliyundrive: source provides GCID, target needs SHA1 → matrix GCID × SHA1 = ❌ cannot rapid upload.
Any source → Google Drive / OneDrive: target has no rapid upload → last column = ➖, normal upload only.

与种子矩阵的关系

Relation to the hash matrix

生成种子时，内容矩阵会依据上述能力预检：若源驱动已提供目标所需全部哈希，则无需下载文件即可生成；否则会下载后边下载边计算。勾选 .torrent 会强制计算 SHA-1 完整 + 分片；勾选 .cas 会强制计算 MD5 完整 + 分片（固定 10 MiB），以满足天翼云秒传。

The hash matrix is preflighted against these capabilities: if the source drive already provides every hash the target needs, no download is required; otherwise files are streamed and hashed on the fly. Selecting .torrent forces SHA-1 whole + pieces; selecting .cas forces MD5 whole + pieces (fixed 10 MiB) for 189pc rapid upload.

说明

Notes

GCID 为分块 SHA1（40 位），用于 PikPak / 迅雷族 / febbox。
天翼云 189 与百度网盘的分片上传另需逐片 MD5（slice_md5），因此矩阵中对应列为 MD5 + 分片，源仅提供 MD5 时记为 🔶（需下载补算分片）。
阿里云盘族使用 pre_hash（文件头 1024 字节 SHA1）触发秒传，命中后补算完整 SHA1。
移动云盘 139 快传使用 SHA256；123 Open 支持 SHA1 复用（sha1_reuse）或 MD5 任选其一。
注意：矩阵反映各驱动的原生秒传能力。当前种子「秒传保存」流程仅天翼云 189pc 的 CAS 秒传被完整接入，其余驱动的秒传在普通上传（Put / PutRapid）路径中生效。

GCID is a block SHA1 (40 hex chars), used by PikPak / Thunder family / febbox.
189 and Baidu chunked uploads additionally need per-piece MD5 (slice_md5), so their column is MD5 + pieces; a source providing only MD5 is marked 🔶 (must download to compute pieces).
The Aliyundrive family uses pre_hash (SHA1 of the first 1024 bytes) to trigger rapid upload, then computes the full SHA1 on hit.
139 rapid upload uses SHA256; 123 Open supports either SHA1 reuse (sha1_reuse) or MD5.
Note: the matrix reflects each drive’s native rapid-upload capability. Currently only 189pc CAS rapid upload is wired into the seed "rapid save" flow; other drives’ rapid upload takes effect in the normal upload path (Put / PutRapid).

Contributors

Edit this page on GitHub

Last updated: