returning cid in ETag instead of MD5 hash

Author: ehsan6sha

.gitignore

@@ -1,3 +1,4 @@
 /target
 .env
 /target/release
+/.claude

crates/fula-cli/src/handlers/multipart.rs

@@ -13,7 +13,6 @@ use axum::{
 use bytes::Bytes;
 use fula_blockstore::{BlockStore, PinStore};
 use fula_core::metadata::ObjectMetadata;
-use fula_crypto::hashing::md5_hash;
 use serde::Deserialize;
 use std::sync::Arc;
 
@@ -116,7 +115,9 @@ pub async fn upload_part(
 
     // Store part data
     let cid = state.block_store.put_block(&body).await?;
-    let etag = md5_hash(&body);
+
+    // Use CID as ETag for parts (content-addressed identifier)
+    let etag = cid.to_string();
 
     let part = UploadPart::new(
         part_number,
@@ -158,13 +159,17 @@ pub async fn complete_multipart_upload(
         return Err(ApiError::s3(S3ErrorCode::InvalidArgument, "Bucket/key mismatch"));
     }
 
-    // Calculate final ETag (MD5 of ETags + "-N")
+    // Calculate final ETag using CIDs (S3 multipart format: {hash}-{partCount})
+    // We hash the concatenated part CIDs to create a unique identifier
     let part_count = upload.parts.len();
-    let mut etag_concat = String::new();
+    let mut cid_concat = String::new();
     for part in upload.sorted_parts() {
-        etag_concat.push_str(&part.etag);
+        cid_concat.push_str(&part.cid);
     }
-    let final_etag = format!("{}-{}", md5_hash(etag_concat.as_bytes()), part_count);
+    // Use BLAKE3 hash of concatenated CIDs, truncated to 32 hex chars for compatibility
+    let hash = fula_crypto::hashing::hash(cid_concat.as_bytes());
+    let hash_hex = hex::encode(&hash.as_bytes()[..16]); // First 16 bytes = 32 hex chars
+    let final_etag = format!("{}-{}", hash_hex, part_count);
 
     // Calculate total size
     let total_size: u64 = upload.parts.values().map(|p| p.size).sum();

crates/fula-cli/src/handlers/object.rs

@@ -32,15 +32,17 @@ pub async fn put_object(
 
     // Store the data
     let cid = state.block_store.put_block(&body).await?;
-    
-    // Calculate ETag (MD5)
-    let etag = md5_hash(&body);
 
-    // Verify Content-MD5 if present
+    // Use CID as ETag (content-addressed identifier)
+    // This is S3-compliant: AWS docs state "The ETag may or may not be an MD5 digest"
+    let etag = cid.to_string();
+
+    // Verify Content-MD5 if present (still uses MD5 for S3 compatibility)
     if let Some(md5_header) = headers.get("Content-MD5").and_then(|v| v.to_str().ok()) {
         if let Ok(expected_bytes) = general_purpose::STANDARD.decode(md5_header) {
             let expected_hex = hex::encode(expected_bytes);
-            if etag != expected_hex {
+            let actual_md5 = md5_hash(&body);
+            if actual_md5 != expected_hex {
                 return Err(ApiError::s3(S3ErrorCode::InvalidDigest, "The Content-MD5 you specified did not match what we received."));
             }
         } else {

crates/fula-cli/src/multipart.rs

@@ -75,7 +75,7 @@ impl MultipartUpload {
 pub struct UploadPart {
     /// Part number (1-10000)
     pub part_number: u32,
-    /// ETag (MD5 hash of part content)
+    /// ETag (CID of part content - content-addressed identifier)
     pub etag: String,
     /// Part size in bytes
     pub size: u64,

crates/fula-core/src/metadata.rs

@@ -63,7 +63,8 @@ pub struct ObjectMetadata {
     /// Size in bytes
     pub size: u64,
 
-    /// ETag (usually MD5 hash for S3 compatibility)
+    /// ETag (CID for single uploads, BLAKE3-hash-of-CIDs-{partCount} for multipart)
+    /// Using CID as ETag is S3-compliant: AWS docs state "The ETag may or may not be an MD5 digest"
     pub etag: String,
 
     /// Last modified timestamp

docs/openAPI.yaml

@@ -268,14 +268,14 @@ paths:
                   <Contents>
                     <Key>photos/beach.jpg</Key>
                     <LastModified>2025-12-07T01:42:00.000Z</LastModified>
-                    <ETag>"b2419b1e3fd45d596ee22bdf62aaaa2f"</ETag>
+                    <ETag>"bafybeigdyrzt5sfp7udm7hu76uh7y26nf3efuylqabf3oc"</ETag>
                     <Size>2048576</Size>
                     <StorageClass>STANDARD</StorageClass>
                   </Contents>
                   <Contents>
                     <Key>photos/sunset.png</Key>
                     <LastModified>2025-12-07T03:15:00.000Z</LastModified>
-                    <ETag>"8f47e10b6a3c9d2e5f1a4b7c3d8e6f9a"</ETag>
+                    <ETag>"bafkreihxvh4d5pqhkhxqfdy3yzpfmam5a6t3wbmtj7d3iur"</ETag>
                     <Size>1536000</Size>
                     <StorageClass>STANDARD</StorageClass>
                   </Contents>
@@ -456,10 +456,13 @@ paths:
           description: Object uploaded successfully
           headers:
             ETag:
-              description: MD5 hash of the object or IPFS CID
+              description: |
+                IPFS CID of the uploaded object (content-addressed identifier).
+                This is S3-compliant as AWS states "The ETag may or may not be an MD5 digest."
+                The CID can be used directly to retrieve the object from IPFS.
               schema:
                 type: string
-              example: '"b2419b1e3fd45d596ee22bdf62aaaa2f"'
+              example: '"bafybeigdyrzt5sfp7udm7hu76uh7y26nf3efuylqabf3oc"'
             x-amz-version-id:
               description: Version ID (if versioning enabled)
               schema:
@@ -838,11 +841,11 @@ paths:
           headers:
             ETag:
               description: |
-                ETag of the uploaded part (IPFS CID).
+                IPFS CID of the uploaded part (content-addressed identifier).
                 **Important**: Save this for CompleteMultipartUpload.
               schema:
                 type: string
-              example: '"a54357aff0632cce46d942af68356b38"'
+              example: '"bafybeigdyrzt5sfp7udm7hu76uh7y26nf3efuylqabf3oc"'
         '400':
           description: Bad Request (part too small, invalid upload ID)
           content:
@@ -894,11 +897,11 @@ paths:
               <CompleteMultipartUpload>
                 <Part>
                   <PartNumber>1</PartNumber>
-                  <ETag>"a54357aff0632cce46d942af68356b38"</ETag>
+                  <ETag>"bafybeigdyrzt5sfp7udm7hu76uh7y26nf3efuylqabf3oc"</ETag>
                 </Part>
                 <Part>
                   <PartNumber>2</PartNumber>
-                  <ETag>"0c78aef83f66abc1fa1e8477f296d394"</ETag>
+                  <ETag>"bafkreihxvh4d5pqhkhxqfdy3yzpfmam5a6t3wbmtj7d3iur"</ETag>
                 </Part>
               </CompleteMultipartUpload>
       responses:
@@ -914,8 +917,9 @@ paths:
                   <Location>https://gateway.example.com/my-bucket/large-file.bin</Location>
                   <Bucket>my-bucket</Bucket>
                   <Key>large-file.bin</Key>
-                  <ETag>"3858f62230ac3c915f300c664312c11f-2"</ETag>
+                  <ETag>"a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6-2"</ETag>
                 </CompleteMultipartUploadResult>
+                # Note: Multipart ETag format is {BLAKE3-hash-of-CIDs}-{partCount}
         '400':
           description: Invalid part list or missing parts
           content:
@@ -1132,7 +1136,7 @@ paths:
                 <?xml version="1.0" encoding="UTF-8"?>
                 <CopyObjectResult xmlns="http://s3.amazonaws.com/doc/2006-03-01/">
                   <LastModified>2025-12-07T12:00:00.000Z</LastModified>
-                  <ETag>"b2419b1e3fd45d596ee22bdf62aaaa2f"</ETag>
+                  <ETag>"bafybeigdyrzt5sfp7udm7hu76uh7y26nf3efuylqabf3oc"</ETag>
                 </CopyObjectResult>
         '404':
           $ref: '#/components/responses/NoSuchKey'