<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	>

<channel>
	<title>Gea-Suan Lin&#039;s BLOG</title>
	<atom:link href="https://blog.gslin.org/feed/" rel="self" type="application/rss+xml" />
	<link>https://blog.gslin.org</link>
	<description>幹壞事是進步最大的原動力</description>
	<lastBuildDate>Sun, 22 Nov 2026 04:54:01 +0000</lastBuildDate>
	<language>en-US</language>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	<generator>https://wordpress.org/?v=7.1.5</generator>
<atom:link rel="hub" href="https://pubsubhubbub.appspot.com"/>
<atom:link rel="hub" href="https://pubsubhubbub.superfeedr.com"/>
<atom:link rel="hub" href="https://websubhub.com/hub"/>
<atom:link rel="self" href="https://blog.gslin.org/feed/"/>
<site xmlns="com-wordpress:feed-additions:1">21326247</site>	<item>
		<title>Amazon Route 53 的加值服務：Advanced Hosted Zones</title>
		<link>https://blog.gslin.org/archives/2026/11/22/13387/amazon-route-53-%e7%9a%84%e5%8a%a0%e5%80%bc%e6%9c%8d%e5%8b%99%ef%bc%9aadvanced-hosted-zones/</link>
					<comments>https://blog.gslin.org/archives/2026/11/22/13387/amazon-route-53-%e7%9a%84%e5%8a%a0%e5%80%bc%e6%9c%8d%e5%8b%99%ef%bc%9aadvanced-hosted-zones/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Sun, 22 Nov 2026 04:54:01 +0000</pubDate>
				<category><![CDATA[AWS]]></category>
		<category><![CDATA[Cloud]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[DNS]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[53]]></category>
		<category><![CDATA[advanced]]></category>
		<category><![CDATA[amazon]]></category>
		<category><![CDATA[aws]]></category>
		<category><![CDATA[cloud]]></category>
		<category><![CDATA[cost]]></category>
		<category><![CDATA[dns]]></category>
		<category><![CDATA[enterprise]]></category>
		<category><![CDATA[hosted]]></category>
		<category><![CDATA[premium]]></category>
		<category><![CDATA[pricing]]></category>
		<category><![CDATA[route]]></category>
		<category><![CDATA[route53]]></category>
		<category><![CDATA[service]]></category>
		<category><![CDATA[zones]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13387</guid>

					<description><![CDATA[看起來 AWS 在 Amazon Route 53 推出加值服務，費用從本來的 US$0.5/mo 或 US$0.1/mo 提升到 US$5/mo，比本來的 zone 貴了 10 倍到 50 倍，看起來是主打企業市場的需求： Amazon Route 53 now offers Advanced Hosted Zones Amazon Route 53 private hosted zones now support public DNS fallback 可以看到蠻多是稽核與防護機制的功能： Amazon Route 53 now offers Advanced Hosted Zones, a new hosted zone tier with additional guardrails, visibility, &#8230; <a href="https://blog.gslin.org/archives/2026/11/22/13387/amazon-route-53-%e7%9a%84%e5%8a%a0%e5%80%bc%e6%9c%8d%e5%8b%99%ef%bc%9aadvanced-hosted-zones/" class="more-link">Continue reading<span class="screen-reader-text"> "Amazon Route 53 的加值服務：Advanced Hosted Zones"</span></a>]]></description>
										<content:encoded><![CDATA[<p>看起來 <a href="https://en.wikipedia.org/wiki/Amazon_Web_Services">AWS</a> 在 <a href="https://en.wikipedia.org/wiki/Amazon_Route_53">Amazon Route 53</a> 推出加值服務，費用從本來的 US$0.5/mo 或 US$0.1/mo 提升到 US$5/mo，比本來的 zone 貴了 10 倍到 50 倍，看起來是主打企業市場的需求：</p>
<ul>
<li><a href="https://aws.amazon.com/about-aws/whats-new/2026/11/route-53-offers-advanced-hosted-zones/">Amazon Route 53 now offers Advanced Hosted Zones</a></li>
<li><a href="https://aws.amazon.com/about-aws/whats-new/2026/11/route-53-public-dns-fallback/">Amazon Route 53 private hosted zones now support public DNS fallback</a></li>
</ul>
<p>可以看到蠻多是稽核與防護機制的功能：</p>
<blockquote><p>Amazon Route 53 now offers Advanced Hosted Zones, a new hosted zone tier with additional guardrails, visibility, and resilience features. Advanced Hosted Zones include Change Guardrails that prevent accidental record deletions, record-level Change History, and public DNS fallback in private hosted zones.</p></blockquote>
<p>就企業來說應該不是問題的定價，先記錄起來...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/22/13387/amazon-route-53-%e7%9a%84%e5%8a%a0%e5%80%bc%e6%9c%8d%e5%8b%99%ef%bc%9aadvanced-hosted-zones/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13387</post-id>	</item>
		<item>
		<title>Amazon S3 對 General Purpose Bucket 的保障讀取速率從 5.5k reqs/sec 提高一倍，到 11k reqs/sec</title>
		<link>https://blog.gslin.org/archives/2026/11/22/13386/amazon-s3-%e5%b0%8d-general-purpose-bucket-%e7%9a%84%e4%bf%9d%e9%9a%9c%e8%ae%80%e5%8f%96%e9%80%9f%e7%8e%87%e5%be%9e-5-5k-reqs-sec-%e6%8f%90%e9%ab%98%e4%b8%80%e5%80%8d%ef%bc%8c%e5%88%b0-11k-reqs-sec/</link>
					<comments>https://blog.gslin.org/archives/2026/11/22/13386/amazon-s3-%e5%b0%8d-general-purpose-bucket-%e7%9a%84%e4%bf%9d%e9%9a%9c%e8%ae%80%e5%8f%96%e9%80%9f%e7%8e%87%e5%be%9e-5-5k-reqs-sec-%e6%8f%90%e9%ab%98%e4%b8%80%e5%80%8d%ef%bc%8c%e5%88%b0-11k-reqs-sec/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Sun, 22 Nov 2026 04:44:54 +0000</pubDate>
				<category><![CDATA[AWS]]></category>
		<category><![CDATA[Cloud]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[amazon]]></category>
		<category><![CDATA[aws]]></category>
		<category><![CDATA[bucket]]></category>
		<category><![CDATA[cloud]]></category>
		<category><![CDATA[general]]></category>
		<category><![CDATA[performance]]></category>
		<category><![CDATA[purpose]]></category>
		<category><![CDATA[rate]]></category>
		<category><![CDATA[request]]></category>
		<category><![CDATA[s3]]></category>
		<category><![CDATA[service]]></category>
		<category><![CDATA[speed]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13386</guid>

					<description><![CDATA[從 AWS 的 feed 看到的：「Amazon S3 increases the read request rate for general purpose buckets by up to 2x」。 這邊都是下限，也就是保障的速率，實際跑通常會高出不少 (當後端的 sharding 打的夠散的時候)： Amazon S3 now supports at least 11,000 GET/HEAD requests per second, up from 5,500, for general purpose buckets. 對於短時間大量存取的應用會縮短時間，像是 batch &#38; 分析類的應用： With this higher request rate, you can retrieve data faster &#8230; <a href="https://blog.gslin.org/archives/2026/11/22/13386/amazon-s3-%e5%b0%8d-general-purpose-bucket-%e7%9a%84%e4%bf%9d%e9%9a%9c%e8%ae%80%e5%8f%96%e9%80%9f%e7%8e%87%e5%be%9e-5-5k-reqs-sec-%e6%8f%90%e9%ab%98%e4%b8%80%e5%80%8d%ef%bc%8c%e5%88%b0-11k-reqs-sec/" class="more-link">Continue reading<span class="screen-reader-text"> "Amazon S3 對 General Purpose Bucket 的保障讀取速率從 5.5k reqs/sec 提高一倍，到 11k reqs/sec"</span></a>]]></description>
										<content:encoded><![CDATA[<p>從 <a href="https://en.wikipedia.org/wiki/Amazon_Web_Services">AWS</a> 的 feed 看到的：「<a href="https://aws.amazon.com/about-aws/whats-new/2026/11/amazon-s3-increases-read-request-rate/">Amazon S3 increases the read request rate for general purpose buckets by up to 2x</a>」。</p>
<p>這邊都是下限，也就是保障的速率，實際跑通常會高出不少 (當後端的 sharding 打的夠散的時候)：</p>
<blockquote><p>Amazon S3 now supports at least 11,000 GET/HEAD requests per second, up from 5,500, for general purpose buckets.</p></blockquote>
<p>對於短時間大量存取的應用會縮短時間，像是 batch &amp; 分析類的應用：</p>
<blockquote><p>With this higher request rate, you can retrieve data faster for read-intensive workloads such as machine learning training, data analytics, and data transfers. To maximize throughput, distribute your requests across multiple prefixes so that S3 can scale each one independently.</p></blockquote>
<p>不過會撞到 rate limit 的人都應該有對應的機制了，這邊應該是還好，主要是讓後來的人 (從暴力硬幹法開始實作了人) 比較不容易撞到？</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/22/13386/amazon-s3-%e5%b0%8d-general-purpose-bucket-%e7%9a%84%e4%bf%9d%e9%9a%9c%e8%ae%80%e5%8f%96%e9%80%9f%e7%8e%87%e5%be%9e-5-5k-reqs-sec-%e6%8f%90%e9%ab%98%e4%b8%80%e5%80%8d%ef%bc%8c%e5%88%b0-11k-reqs-sec/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13386</post-id>	</item>
		<item>
		<title>可以跟 Qwen3.8-27B 打的 Inkling 2 Mini</title>
		<link>https://blog.gslin.org/archives/2026/11/21/13385/%e5%8f%af%e4%bb%a5%e8%b7%9f-qwen3-8-27b-%e6%89%93%e7%9a%84-inkling-2-mini/</link>
					<comments>https://blog.gslin.org/archives/2026/11/21/13385/%e5%8f%af%e4%bb%a5%e8%b7%9f-qwen3-8-27b-%e6%89%93%e7%9a%84-inkling-2-mini/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Fri, 20 Nov 2026 21:53:05 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[2]]></category>
		<category><![CDATA[ai]]></category>
		<category><![CDATA[inkling]]></category>
		<category><![CDATA[language]]></category>
		<category><![CDATA[large]]></category>
		<category><![CDATA[llm]]></category>
		<category><![CDATA[machines]]></category>
		<category><![CDATA[mini]]></category>
		<category><![CDATA[model]]></category>
		<category><![CDATA[moe]]></category>
		<category><![CDATA[thinking]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13385</guid>

					<description><![CDATA[Thinking Machines 推出新的模型 Inkling 2 Mini，可以跟 Qwen3.8-27B 對打的等級：「Introducing Inkling 2 Mini (via)」。 首先是目前大家比較喜歡拿來用的 Artificial Analysis (AA)，在 AA 上的分數就跟 Qwen3.8-27B 相同了，都是 20 分： 然後居然是 MoE 模型打 Qwen3.8-27B，用 5.4B/26.4B 對上 dense 的 Qwen3.8-27B，這競爭力很強啊... 回台灣後應該也有很多人轉好不同格式了，到時候來跑看看 :o]]></description>
										<content:encoded><![CDATA[<p><a href="https://thinkingmachines.ai/">Thinking Machines</a> 推出新的模型 Inkling 2 Mini，可以跟 <a href="https://huggingface.co/Qwen/Qwen3.8-27B">Qwen3.8-27B</a> 對打的等級：「<a href="https://thinkingmachines.ai/news/inkling-2-mini/">Introducing Inkling 2 Mini</a> (<a href="https://news.ycombinator.com/item?id=50581260">via</a>)」。</p>
<p>首先是目前大家比較喜歡拿來用的 <a href="https://artificialanalysis.ai/">Artificial Analysis</a> (AA)，在 AA 上的分數就跟 Qwen3.8-27B 相同了，都是 20 分：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1795211158-fab1c7e5.webp" /><img decoding="async" src="https://i.gslin.com/s/1795211158-fab1c7e5.png" alt="" /></picture>
<p>然後居然是 MoE 模型打 Qwen3.8-27B，用 5.4B/26.4B 對上 dense 的 Qwen3.8-27B，這競爭力很強啊... 回台灣後應該也有很多人轉好不同格式了，到時候來跑看看 :o</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/21/13385/%e5%8f%af%e4%bb%a5%e8%b7%9f-qwen3-8-27b-%e6%89%93%e7%9a%84-inkling-2-mini/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13385</post-id>	</item>
		<item>
		<title>Amazon EC2 推出 T9g</title>
		<link>https://blog.gslin.org/archives/2026/11/21/13384/amazon-ec2-%e6%8e%a8%e5%87%ba-t9g/</link>
					<comments>https://blog.gslin.org/archives/2026/11/21/13384/amazon-ec2-%e6%8e%a8%e5%87%ba-t9g/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Fri, 20 Nov 2026 21:42:49 +0000</pubDate>
				<category><![CDATA[AWS]]></category>
		<category><![CDATA[Cloud]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Hardware]]></category>
		<category><![CDATA[Infrastructure]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[amazon]]></category>
		<category><![CDATA[arm]]></category>
		<category><![CDATA[aws]]></category>
		<category><![CDATA[cloud]]></category>
		<category><![CDATA[ec2]]></category>
		<category><![CDATA[family]]></category>
		<category><![CDATA[instance]]></category>
		<category><![CDATA[service]]></category>
		<category><![CDATA[t6g]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13384</guid>

					<description><![CDATA[AWS 在 ARM 系列的 burstable 產品線更新了，推出了 T9g 的產品線：「Introducing Amazon EC2 T9g instances based on Graviton5」。 官方宣稱的效能是原來的 2.3 倍： T9g instances deliver up to 2.3x the compute performance, 1.5x higher network bandwidth, and 2.8x higher EBS bandwidth compared to T4g instances. 拿 us-east-1 的價錢來看，t4g.small 是 $0.0168/hr，t9g.small 是 $0.02083/hr，大約貴了 24%，宣稱整體 45% 的 CP 值提升不知道怎麼算出來的： T9g instances deliver &#8230; <a href="https://blog.gslin.org/archives/2026/11/21/13384/amazon-ec2-%e6%8e%a8%e5%87%ba-t9g/" class="more-link">Continue reading<span class="screen-reader-text"> "Amazon EC2 推出 T9g"</span></a>]]></description>
										<content:encoded><![CDATA[<p><a href="https://en.wikipedia.org/wiki/Amazon_Web_Services">AWS</a> 在 <a href="https://en.wikipedia.org/wiki/ARM_architecture_family">ARM</a> 系列的 burstable 產品線更新了，推出了 T9g 的產品線：「<a href="https://aws.amazon.com/about-aws/whats-new/2026/11/ec2-t9g-instances-ga/">Introducing Amazon EC2 T9g instances based on Graviton5</a>」。</p>
<p>官方宣稱的效能是原來的 2.3 倍：</p>
<blockquote><p>T9g instances deliver up to 2.3x the compute performance, 1.5x higher network bandwidth, and 2.8x higher EBS bandwidth compared to T4g instances.</p></blockquote>
<p>拿 <code>us-east-1</code> 的價錢來看，<code>t4g.small</code> 是 $0.0168/hr，<code>t9g.small</code> 是 $0.02083/hr，大約貴了 24%，宣稱整體 45% 的 CP 值提升不知道怎麼算出來的：</p>
<blockquote><p>T9g instances deliver up to 45% better price performance compared to previous generation T4g instances, and offer four sizes: nano, micro, small, and medium.</p></blockquote>
<p>另外比較特別的是只提供四個 size，最多到 4GB RAM 的機種，查了一下當初的 <code>t8i</code> 也是一樣的設計了，對於需要比較大記憶體的人就得跳去用 <code>m</code>/<code>c</code>/<code>r</code> 系列的機種了？</p>
<p>anyway，馬上想到的是，對於當初因為 CPU 用比較多而升級的應用，可以考慮降一級來跑 <code>t9g</code>。</p>
<p>目前支援的區域是這些，目前還沒看到 <code>ap-east-2</code> 台北區，但常見的傳統重點區域看起來都上了：</p>
<blockquote><p>T9g instances are available in US East (N. Virginia, Ohio), US West (N. California, Oregon), Canada (Central), South America (São Paulo), Europe (Frankfurt, Spain), and Asia Pacific (Malaysia, Singapore, Sydney, Tokyo) regions.</p></blockquote>
<p>因為還是有些價差，剛開始完全還沒有量的時候用 <code>t4g</code>，有量以後再看看是 CPU 吃比較重，還是 memory 需求比較高，來決定要不要上 <code>t9g</code> 以及後續的 RI 或 savings plan 的購買...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/21/13384/amazon-ec2-%e6%8e%a8%e5%87%ba-t9g/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13384</post-id>	</item>
		<item>
		<title>Valkey 9.2 改了/加了不少東西</title>
		<link>https://blog.gslin.org/archives/2026/11/20/13383/valkey-9-2-%e6%94%b9%e4%ba%86-%e5%8a%a0%e4%ba%86%e4%b8%8d%e5%b0%91%e6%9d%b1%e8%a5%bf/</link>
					<comments>https://blog.gslin.org/archives/2026/11/20/13383/valkey-9-2-%e6%94%b9%e4%ba%86-%e5%8a%a0%e4%ba%86%e4%b8%8d%e5%b0%91%e6%9d%b1%e8%a5%bf/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Fri, 20 Nov 2026 08:43:30 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Programming]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[b]]></category>
		<category><![CDATA[compression]]></category>
		<category><![CDATA[data]]></category>
		<category><![CDATA[license]]></category>
		<category><![CDATA[list]]></category>
		<category><![CDATA[lz4]]></category>
		<category><![CDATA[redis]]></category>
		<category><![CDATA[replication]]></category>
		<category><![CDATA[set]]></category>
		<category><![CDATA[skip]]></category>
		<category><![CDATA[sorted]]></category>
		<category><![CDATA[structure]]></category>
		<category><![CDATA[tree]]></category>
		<category><![CDATA[valkey]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13383</guid>

					<description><![CDATA[Valkey 從 Redis 改 license 後 fork 出來，然後最近這版丟出了有趣的東西：「Valkey 9.2 offers forkless snapshots, compressed replication, ACL roles, and more」。 首先是 replication 支援壓縮，目前支援 LZ4： The new repl-compression configuration adds streaming compression for steady-state replication, negotiated per replica so older or opted-out replicas keep receiving plaintext (#3853). In the PR's benchmark of 3 million keys of about 315 &#8230; <a href="https://blog.gslin.org/archives/2026/11/20/13383/valkey-9-2-%e6%94%b9%e4%ba%86-%e5%8a%a0%e4%ba%86%e4%b8%8d%e5%b0%91%e6%9d%b1%e8%a5%bf/" class="more-link">Continue reading<span class="screen-reader-text"> "Valkey 9.2 改了/加了不少東西"</span></a>]]></description>
										<content:encoded><![CDATA[<p><a href="https://en.wikipedia.org/wiki/Valkey">Valkey</a> 從 <a href="https://en.wikipedia.org/wiki/Redis">Redis</a> 改 license 後 fork 出來，然後最近這版丟出了有趣的東西：「<a href="https://valkey.io/blog/valkey-9.2-release/">Valkey 9.2 offers forkless snapshots, compressed replication, ACL roles, and more</a>」。</p>
<p>首先是 replication 支援壓縮，目前支援 <a href="https://en.wikipedia.org/wiki/LZ4_(compression_algorithm)">LZ4</a>：</p>
<blockquote><p>The new repl-compression configuration adds streaming compression for steady-state replication, negotiated per replica so older or opted-out replicas keep receiving plaintext (#3853). In the PR's benchmark of 3 million keys of about 315 bytes each, LZ4 cut replication bandwidth by 52% with less than 1% throughput overhead.</p></blockquote>
<p>這讓我想到之前 server 主流還是 1Gbps 的網卡的年代，當時發現 request 速度變慢，一路追才發現 <a href="https://en.wikipedia.org/wiki/Memcached">memcached</a> server 的網路的事情...</p>
<p>另外是 sorted set 換資料結構了，從 <a href="https://en.wikipedia.org/wiki/Skip_list">skip list</a> 換成 <a href="https://en.wikipedia.org/wiki/B%2B_tree">B+ tree</a>，大幅降低了 overhead 的部分：</p>
<blockquote><p>Large sorted sets are now backed by a B+ tree instead of a skiplist (#4359). The PR reports per-member memory overhead dropping from 50.3 bytes to between 28.5 and 32.0 bytes, 36% to 43% less, for 5 million members with 20-byte elements. The B+ tree also compacts sparse leaves when members are deleted, so a sorted set that shrinks or has frequent score updates gives memory back.</p></blockquote>
<p>整體看起來改了好多東西，整個開發能量比 Redis 高不少...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/20/13383/valkey-9-2-%e6%94%b9%e4%ba%86-%e5%8a%a0%e4%ba%86%e4%b8%8d%e5%b0%91%e6%9d%b1%e8%a5%bf/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13383</post-id>	</item>
		<item>
		<title>Ruby 的 RegEx 也有 RCE 問題</title>
		<link>https://blog.gslin.org/archives/2026/11/19/13382/ruby-%e7%9a%84-regex-%e4%b9%9f%e6%9c%89-rce-%e5%95%8f%e9%a1%8c/</link>
					<comments>https://blog.gslin.org/archives/2026/11/19/13382/ruby-%e7%9a%84-regex-%e4%b9%9f%e6%9c%89-rce-%e5%95%8f%e9%a1%8c/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Thu, 19 Nov 2026 00:55:41 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Programming]]></category>
		<category><![CDATA[Security]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[advisory]]></category>
		<category><![CDATA[ai]]></category>
		<category><![CDATA[code]]></category>
		<category><![CDATA[engine]]></category>
		<category><![CDATA[execution]]></category>
		<category><![CDATA[expression]]></category>
		<category><![CDATA[language]]></category>
		<category><![CDATA[large]]></category>
		<category><![CDATA[llm]]></category>
		<category><![CDATA[model]]></category>
		<category><![CDATA[rce]]></category>
		<category><![CDATA[regex]]></category>
		<category><![CDATA[regular]]></category>
		<category><![CDATA[remote]]></category>
		<category><![CDATA[ruby]]></category>
		<category><![CDATA[security]]></category>
		<category><![CDATA[vulnerability]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13382</guid>

					<description><![CDATA[在「Security advisories: multiple vulnerabilities in Regexp and String case conversion (via)」這邊看到 Ruby 也發公告了： Seven vulnerabilities have been discovered in Ruby’s regular expression engine and in the case conversion of strings. We are aware of PoCs that prove some of them can be used to trigger remote code execution. 這七個裡面，有五個官方說可以 RCE。 Hacker News 的 id=50554474 說的很好： &#8230; <a href="https://blog.gslin.org/archives/2026/11/19/13382/ruby-%e7%9a%84-regex-%e4%b9%9f%e6%9c%89-rce-%e5%95%8f%e9%a1%8c/" class="more-link">Continue reading<span class="screen-reader-text"> "Ruby 的 RegEx 也有 RCE 問題"</span></a>]]></description>
										<content:encoded><![CDATA[<p>在「<a href="https://www.ruby-lang.org/en/news/2026/11/18/multiple-vulnerabilities-in-regexp-and-string/">Security advisories: multiple vulnerabilities in Regexp and String case conversion</a> (<a href="https://news.ycombinator.com/item?id=50554038">via</a>)」這邊看到 <a href="https://en.wikipedia.org/wiki/Ruby_(programming_language)">Ruby</a> 也發公告了：</p>
<blockquote><p>Seven vulnerabilities have been discovered in Ruby’s regular expression engine and in the case conversion of strings. We are aware of PoCs that prove some of them can be used to trigger remote code execution.</p></blockquote>
<p>這七個裡面，有五個官方說可以 <a href="https://en.wikipedia.org/wiki/Remote_code_execution">RCE</a>。</p>
<p><a href="https://en.wikipedia.org/wiki/Hacker_News">Hacker News</a> 的 <a href="https://news.ycombinator.com/item?id=50554474">id=50554474</a> 說的很好：</p>
<blockquote><p>I need a vacation</p></blockquote>
<p>最近這波專打程式語言的字串處理，看起來都被找到不少東西...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/19/13382/ruby-%e7%9a%84-regex-%e4%b9%9f%e6%9c%89-rce-%e5%95%8f%e9%a1%8c/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13382</post-id>	</item>
		<item>
		<title>Java 的 Base64 的處理問題</title>
		<link>https://blog.gslin.org/archives/2026/11/19/13381/java-%e7%9a%84-base64-%e7%9a%84%e8%99%95%e7%90%86%e5%95%8f%e9%a1%8c/</link>
					<comments>https://blog.gslin.org/archives/2026/11/19/13381/java-%e7%9a%84-base64-%e7%9a%84%e8%99%95%e7%90%86%e5%95%8f%e9%a1%8c/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Thu, 19 Nov 2026 00:28:33 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Hardware]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Programming]]></category>
		<category><![CDATA[Security]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[advisory]]></category>
		<category><![CDATA[ai]]></category>
		<category><![CDATA[assembly]]></category>
		<category><![CDATA[avx-512]]></category>
		<category><![CDATA[base64]]></category>
		<category><![CDATA[issue]]></category>
		<category><![CDATA[java]]></category>
		<category><![CDATA[language]]></category>
		<category><![CDATA[large]]></category>
		<category><![CDATA[llm]]></category>
		<category><![CDATA[model]]></category>
		<category><![CDATA[security]]></category>
		<category><![CDATA[vulnerability]]></category>
		<category><![CDATA[x86-64]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13381</guid>

					<description><![CDATA[Java 這邊也出現很爆炸性的安全漏洞了，這次是對 Base64 的處理：「Java4Shell: Write once, RCE everywhere (via)」。 To decode Base64 faster, OpenJDK uses specialized x86-64 CPU instructions to process many bytes at once. The corruption itself can be triggered by just 64 A characters followed by a single =: AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA= 主因是透過 AVX-512 加速的 asm code 有問題，所以這次的問題僅限於新「一點」的 x86-64： To speed up Base64 decoding, OpenJDK &#8230; <a href="https://blog.gslin.org/archives/2026/11/19/13381/java-%e7%9a%84-base64-%e7%9a%84%e8%99%95%e7%90%86%e5%95%8f%e9%a1%8c/" class="more-link">Continue reading<span class="screen-reader-text"> "Java 的 Base64 的處理問題"</span></a>]]></description>
										<content:encoded><![CDATA[<p><a href="https://en.wikipedia.org/wiki/Java_(programming_language)">Java</a> 這邊也出現很爆炸性的安全漏洞了，這次是對 <a href="https://en.wikipedia.org/wiki/Base64">Base64</a> 的處理：「<a href="https://calif.io/java4shell">Java4Shell: Write once, RCE everywhere</a> (<a href="https://news.ycombinator.com/item?id=50538240">via</a>)」。</p>
<blockquote><p>To decode Base64 faster, OpenJDK uses specialized x86-64 CPU instructions to process many bytes at once.</p></blockquote>
<blockquote><p>The corruption itself can be triggered by just 64 A characters followed by a single =:</p>
<p>AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=</p></blockquote>
<p>主因是透過 <a href="https://en.wikipedia.org/wiki/AVX-512">AVX-512</a> 加速的 asm code 有問題，所以這次的問題僅限於新「一點」的 <a href="https://en.wikipedia.org/wiki/X86-64">x86-64</a>：</p>
<blockquote><p>To speed up Base64 decoding, OpenJDK uses specialized AVX-512 VBMI instructions on supported x86-64 processors. The bug sits in the hand-rolled assembly code that handles the final few characters.</p></blockquote>
<p>官方看起來已經丟出新版了，但如果暫時無法升級的話，這篇建議的 mitigation 有兩條，第一條是關掉 Base64 這邊的指令加速，第二條是直接限制在 <a href="https://en.wikipedia.org/wiki/AVX2">AVX2</a> 上 (不使用 AVX-512)：</p>
<ul>
<li><code>-XX:-UseBASE64Intrinsics</code> to disable hardware-accelerated Base64 encoding and decoding, or</li>
<li><code>-XX:UseAVX=2</code> to disable AVX-512. Apply this flag only on x86-64 hosts. OpenJDK on ARM doesn't recognize it and will fail to start.</li>
</ul>
<p>另外文章有提到現在透過 <a href="https://en.wikipedia.org/wiki/Large_language_model">LLM</a> 加速的關係，整個速度比起以前快太多了：</p>
<blockquote><p>It took our own Quang Luong an hour to discover the vulnerability, and two more hours to develop the first working exploit. The total token cost was less than $200.</p></blockquote>
<p>業界看起來還會有蠻長時間的陣痛期 (好像也不能叫陣痛了...)，尤其當 LLM 愈來愈強，應該就會抓出更多複雜的 security issue...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/19/13381/java-%e7%9a%84-base64-%e7%9a%84%e8%99%95%e7%90%86%e5%95%8f%e9%a1%8c/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13381</post-id>	</item>
		<item>
		<title>又有提前告知將會有 PHP 安全性更新了</title>
		<link>https://blog.gslin.org/archives/2026/11/19/13380/%e5%8f%88%e6%9c%89%e6%8f%90%e5%89%8d%e5%91%8a%e7%9f%a5%e5%b0%87%e6%9c%83%e6%9c%89-php-%e5%ae%89%e5%85%a8%e6%80%a7%e6%9b%b4%e6%96%b0%e4%ba%86/</link>
					<comments>https://blog.gslin.org/archives/2026/11/19/13380/%e5%8f%88%e6%9c%89%e6%8f%90%e5%89%8d%e5%91%8a%e7%9f%a5%e5%b0%87%e6%9c%83%e6%9c%89-php-%e5%ae%89%e5%85%a8%e6%80%a7%e6%9b%b4%e6%96%b0%e4%ba%86/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Wed, 18 Nov 2026 22:26:58 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Programming]]></category>
		<category><![CDATA[Security]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[advance]]></category>
		<category><![CDATA[advisory]]></category>
		<category><![CDATA[critical]]></category>
		<category><![CDATA[in]]></category>
		<category><![CDATA[php]]></category>
		<category><![CDATA[security]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13380</guid>

					<description><![CDATA[在「Upcoming Critical PHP Security Release」這邊： The PHP security team announces upcoming security releases scheduled for Thursday, 19 November 2026, between 14:00 and 14:30 UTC. These releases are part of PHP’s regular four-week release cycle and will address security vulnerabilities affecting all supported PHP versions. The highest severity is CRITICAL. The issues rated CRITICAL require either &#8230; <a href="https://blog.gslin.org/archives/2026/11/19/13380/%e5%8f%88%e6%9c%89%e6%8f%90%e5%89%8d%e5%91%8a%e7%9f%a5%e5%b0%87%e6%9c%83%e6%9c%89-php-%e5%ae%89%e5%85%a8%e6%80%a7%e6%9b%b4%e6%96%b0%e4%ba%86/" class="more-link">Continue reading<span class="screen-reader-text"> "又有提前告知將會有 PHP 安全性更新了"</span></a>]]></description>
										<content:encoded><![CDATA[<p>在「<a href="https://www.php.net/archive/2026.php#2026-11-18-1">Upcoming Critical PHP Security Release</a>」這邊：</p>
<blockquote><p>The PHP security team announces upcoming security releases <strong>scheduled for Thursday, 19 November 2026, between 14:00 and 14:30 UTC</strong>.</p>
<p>These releases are part of PHP’s regular four-week release cycle and will address security vulnerabilities affecting all supported PHP versions. The highest severity is <strong>CRITICAL</strong>.</p>
<p>The issues rated CRITICAL require either non-default configuration or uncommon setups.</p>
<p>Users are strongly encouraged to prepare to upgrade promptly once the releases become available. Technical details and fixed version numbers will be published alongside the releases.</p>
<p>Thank you for helping us keep PHP secure.</p></blockquote>
<p>換算台灣的時間是今天 10:00pm～10:30pm，這次的 CRITICAL 是有動到特殊設定時才會觸發？看起來就等吧...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/19/13380/%e5%8f%88%e6%9c%89%e6%8f%90%e5%89%8d%e5%91%8a%e7%9f%a5%e5%b0%87%e6%9c%83%e6%9c%89-php-%e5%ae%89%e5%85%a8%e6%80%a7%e6%9b%b4%e6%96%b0%e4%ba%86/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13380</post-id>	</item>
		<item>
		<title>Gemini 4 Neon</title>
		<link>https://blog.gslin.org/archives/2026/11/19/13379/gemini-4-neon/</link>
					<comments>https://blog.gslin.org/archives/2026/11/19/13379/gemini-4-neon/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Wed, 18 Nov 2026 22:21:18 +0000</pubDate>
				<category><![CDATA[API]]></category>
		<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[4]]></category>
		<category><![CDATA[ai]]></category>
		<category><![CDATA[gemini]]></category>
		<category><![CDATA[google]]></category>
		<category><![CDATA[language]]></category>
		<category><![CDATA[large]]></category>
		<category><![CDATA[llm]]></category>
		<category><![CDATA[model]]></category>
		<category><![CDATA[neon]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13379</guid>

					<description><![CDATA[看到 Google 又推出新的模型，Gemini 4 Neon：「Introducing Gemini 4 Neon (via)」。 一如往常的用元素命名，這次是 Neon (氖)，但還是看不出來命名規則。 依照說明是目前 Gemini 4 系列最強的 frontier model，Artificial Analysis 上面也給了 61 分，比先前的 Carbon 的 60 分還高一點點，官方 benchmark 時對到的對象是 Opus 5.5 與 GPT-6.2 Sol： 看了一下跟 Carbon 的差距有點小，但比 Carbon 貴了 50%，看起來邊際效應有點重...]]></description>
										<content:encoded><![CDATA[<p>看到 <a href="https://en.wikipedia.org/wiki/Google">Google</a> 又推出新的模型，<a href="https://en.wikipedia.org/wiki/Gemini_(language_model)">Gemini</a> 4 Neon：「<a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/">Introducing Gemini 4 Neon</a> (<a href="https://news.ycombinator.com/item?id=50552424">via</a>)」。</p>
<p>一如往常的用元素命名，這次是 Neon (氖)，但還是看不出來命名規則。</p>
<p>依照說明是目前 Gemini 4 系列最強的 frontier model，<a href="https://artificialanalysis.ai/">Artificial Analysis</a> 上面也給了 61 分，比先前的 Carbon 的 60 分還高一點點，官方 benchmark 時對到的對象是 Opus 5.5 與 GPT-6.2 Sol：</p>
<picture><source type="image/webp" srcset="https://i.gslin.com/s/1795039747-b940460a.webp" /><img decoding="async" src="https://i.gslin.com/s/1795039747-b940460a.png" alt="" /></picture>
<p>看了一下跟 Carbon 的差距有點小，但比 Carbon 貴了 50%，看起來邊際效應有點重...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/19/13379/gemini-4-neon/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13379</post-id>	</item>
		<item>
		<title>nginx 的 RCE</title>
		<link>https://blog.gslin.org/archives/2026/11/19/13376/nginx-%e7%9a%84-rce/</link>
					<comments>https://blog.gslin.org/archives/2026/11/19/13376/nginx-%e7%9a%84-rce/#respond</comments>
		
		<dc:creator><![CDATA[Gea-Suan Lin]]></dc:creator>
		<pubDate>Wed, 18 Nov 2026 16:43:02 +0000</pubDate>
				<category><![CDATA[Computer]]></category>
		<category><![CDATA[Murmuring]]></category>
		<category><![CDATA[Network]]></category>
		<category><![CDATA[Security]]></category>
		<category><![CDATA[Service]]></category>
		<category><![CDATA[Software]]></category>
		<category><![CDATA[WWW]]></category>
		<category><![CDATA[advisory]]></category>
		<category><![CDATA[code]]></category>
		<category><![CDATA[execution]]></category>
		<category><![CDATA[exploit]]></category>
		<category><![CDATA[f5]]></category>
		<category><![CDATA[http/2]]></category>
		<category><![CDATA[mitigation]]></category>
		<category><![CDATA[nginx]]></category>
		<category><![CDATA[rce]]></category>
		<category><![CDATA[remote]]></category>
		<category><![CDATA[security]]></category>
		<category><![CDATA[server]]></category>
		<category><![CDATA[web]]></category>
		<category><![CDATA[www]]></category>
		<guid isPermaLink="false">https://blog.gslin.org/?p=13376</guid>

					<description><![CDATA[nginx 的 RCE：「Unauthenticated Remote Code Execution in NGINX: CVE-2026-123633 and CVE-2026-123007 (via)」。 引用的文章是 Wiz 發表的 (年初的時候被 Alphabet 收購)，但已知有五個團隊都找到一樣的洞： Wiz Research, Ikotas Labs, Metnir, Rapid7, and Xint independently uncovered two critical remote code execution vulnerabilities in NGINX, software used to receive and route traffic to production applications and APIs. Wiz 看起來是第二個 (至少不是第一個) 發現的： October 21, 2026: &#8230; <a href="https://blog.gslin.org/archives/2026/11/19/13376/nginx-%e7%9a%84-rce/" class="more-link">Continue reading<span class="screen-reader-text"> "nginx 的 RCE"</span></a>]]></description>
										<content:encoded><![CDATA[<p><a href="https://en.wikipedia.org/wiki/Nginx">nginx</a> 的 <a href="https://en.wikipedia.org/wiki/Remote_code_execution">RCE</a>：「<a href="https://www.wiz.io/blog/nginx-http2-rce-vulnerabilities">Unauthenticated Remote Code Execution in NGINX: CVE-2026-123633 and CVE-2026-123007</a> (<a href="https://news.ycombinator.com/item?id=50547374">via</a>)」。</p>
<p>引用的文章是 <a href="https://en.wikipedia.org/wiki/Wiz,_Inc.">Wiz</a> 發表的 (年初的時候被 <a href="https://en.wikipedia.org/wiki/Alphabet_Inc.">Alphabet</a> 收購)，但已知有五個團隊都找到一樣的洞：</p>
<blockquote><p>Wiz Research, Ikotas Labs, Metnir, Rapid7, and Xint independently uncovered two critical remote code execution vulnerabilities in NGINX, software used to receive and route traffic to production applications and APIs.</p></blockquote>
<p>Wiz 看起來是第二個 (至少不是第一個) 發現的：</p>
<blockquote><p>October 21, 2026: Wiz reports the HTTP/2 use-after-free vulnerability, later assigned CVE-2026-123633, to F5, including confirmed remote code execution with an ASLR bypass and a proposed fix. F5 confirms that another research team had previously reported the same vulnerability.</p></blockquote>
<p>兩個 CVE 都是 HTTP/2 相關的：</p>
<ul>
<li>HTTP/2 connection use-after-free</li>
<li>HTTP/2 send_window integer underflow</li>
</ul>
<p>這次很慘的是，在發表的當下 <a href="https://en.wikipedia.org/wiki/Debian">Debian</a> 與 <a href="https://en.wikipedia.org/wiki/Ubuntu">Ubuntu</a> 都沒有修正啊，只能選擇關 <a href="https://en.wikipedia.org/wiki/HTTP%2F2">HTTP/2</a> 當作 mitigation 來用了。</p>
<p>另外痛的是，如果是五個團隊都要發現這個問題，就代表大概有六七八九十的團隊有找到一樣的問題但沒有公開... (幹)</p>
<p>雖然因為個人的偏好都跳 <a href="https://en.wikipedia.org/wiki/Caddy_(web_server)">Caddy</a> 跳的差不多了，但還是有些機器因為懶得動跑 nginx，這次就被揍了...</p>
]]></content:encoded>
					
					<wfw:commentRss>https://blog.gslin.org/archives/2026/11/19/13376/nginx-%e7%9a%84-rce/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
		<post-id xmlns="com-wordpress:feed-additions:1">13376</post-id>	</item>
	</channel>
</rss>

<!--
Performance optimized by W3 Total Cache. Learn more: https://www.boldgrid.com/w3-total-cache/?utm_source=w3tc&utm_medium=footer_comment&utm_campaign=free_plugin

Object Caching 40/51 objects using APC
Page Caching using APC (Page is feed) 
Minified using APC
Database Caching using APC

Served from: blog.gslin.org @ 2026-11-22 13:59:49 by W3 Total Cache
-->