After CORS (12 Oct), this post covers change data capture: log-based sources, schema evolution, idempotent sinks, outbox contrast, bootstrap/lag, privacy filters, ordering and deletes.

परिवर्तनप्रवाहः: A Metrical Codification of Change Data Capture (with English glosses)

What this post is about

CDC publishes database row changes as an event stream for search warehouses caches and integrations.

Topics covered

1. Polling pain
2. WAL/binlog capture
3. Schema evolution with consumers
4. At-least-once + idempotent sinks
5. Outbox vs CDC
6. Snapshot bootstrap and lag
7. PII filtering
8. Per-key ordering
9. Deletes/tombstones
10. Links to streams and warehouse

Sanskrit title

परिवर्तनप्रवाहः = parivartana-pravāhaḥ “flow of changes” (CDC).

English

Minimizable; Sanskrit-first.

ग्राह्यम् / Keep

१. log-based CDC: prefer WAL/binlog
२. schema expand: consumers compatible
३. idempotent sinks: PK/event id upserts
४. lag SLO: monitor streaming delay
५. PII filter: column allow-lists
६. bootstrap plan: snapshot then stream
७. delete policy: tombstones explicit

त्याज्यम् / Avoid

• poll whole tables every second as architecture
• fan out raw PII to all sinks
• assume global total order

पदच्छेदः
DB सत्यं यदि केवलं पठने तदा घटनाः विलम्बन्ते । परिवर्तनप्रवाहस्तस्माद् लेखजन्मश्रुतिः ॥१॥

Word-for-word

Gloss table

English sense
Change data capture streams row-level inserts updates and deletes from the database log so downstream systems learn changes without heavy polling.

Context / topic
Topic: why CDC.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
WAL binlog स्रोतं प्रामाणिकम् । trigger exclusively न सर्वदा श्रेयः ॥२॥

Word-for-word

Gloss table

English sense
Log-based CDC usually beats trigger-only designs for completeness and load though triggers still appear in some systems.

Context / topic
Topic: log-based capture.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
schema evolution ग्राहके सह । अन्धconsumers भग्नाः ॥३॥

Word-for-word

Gloss table

English sense
Column adds/renames must be expand/contract compatible with CDC consumers. Breaking payload shapes strand search indexes and warehouses.

Context / topic
Topic: schema and CDC. Links to migrations and API versioning ideas.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
exactly-once sink दुःसाध्यं । इदमर्थलेखनं व्यावहारिकम् ॥४॥

Word-for-word

Gloss table

English sense
Treat delivery as at-least-once and make sinks idempotent by primary key or event id rather than promising magic exactly-once everywhere.

Context / topic
Topic: delivery semantics. Links to idempotency and outbox.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
बहिर्पटलं app जन्म CDC db जन्म । उभौ मार्गौ कदाचित् सह ॥५॥

Word-for-word

Gloss table

English sense
Outbox is application-chosen events; CDC is database truth. Some architectures use both; do not double-publish the same business fact without care.

Context / topic
Topic: outbox vs CDC.

सन्धिः: पटलः सङ्कल्पः; CDC तथ्यम्।

वृत्तमिति: एकादशाक्षराः पादाः।

पदच्छेदः
lag माप्यः snapshot+incremental क्रमः । अन्धbackfill विपत्तिः ॥६॥

Word-for-word

Gloss table

English sense
Bootstrap with a snapshot then switch to the stream at a tracked position. Measure CDC lag like replica lag.

Context / topic
Topic: bootstrap and lag.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
PII स्तम्भाः मा सर्वत्र प्रवाहे । गोप्यताfilter अनिवार्यः ॥७॥

Word-for-word

Gloss table

English sense
Do not fan sensitive columns to every downstream by default. Filter mask or tokenize per consumer need.

Context / topic
Topic: privacy on the stream. Links to privacy logging.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
ordering per key partition । विश्वक्रमः मा मिथ्याप्रतिज्ञा ॥८॥

Word-for-word

Gloss table

English sense
CDC usually preserves order per primary key or partition not a single global total order across all tables.

Context / topic
Topic: ordering guarantees.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
downstream delete tombstone धर्मः । soft delete विवेकः ॥९॥

Word-for-word

Gloss table

English sense
Propagate deletes intentionally. Soft deletes upstream become update events; hard deletes need tombstones consumers understand.

Context / topic
Topic: deletes.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
घटनाप्रवाहः वाहनं CDC स्रोतं । दत्तांशभण्डारः एकः ग्राहकः ॥१०॥

Word-for-word

Gloss table

English sense
Streaming platforms carry change events; warehouses and search are common sinks. CDC is the source pattern that feeds them.

Context / topic
Topic: series links.

सन्धिः: CDC जायते; प्रवाहः वहति; भण्डारः पचति।

वृत्तमिति: एकादशाक्षराः पादाः।

पदच्छेदः
स्रोतः schema lag गोप्यता sink-इदमर्थः पञ्चकम् । एभिः परिवर्तनप्रवाहः लेखसत्यसमन्वितः ॥११॥

Word-for-word

Gloss table

English sense
Five limbs: log-based source, schema discipline, lag metrics, privacy filters and idempotent sinks.

Context / topic
Topic: checklist.

वृत्तमिति: ८-८-८-८।

पदच्छेदः
इति परिवर्तनप्रवाहसारो लेखसत्यसमन्वितः । मितं श्रवणं क्रियायोग्यं तन्त्रधर्मः स उच्यते ॥१२॥

Word-for-word

Gloss table

English sense
Closing: hear the database truth as a stream not as endless full-table polls.

Context / topic
Topic: close.

उपसंहारन्यायः: लॉगात् शृणु; lag मापय; sink इदमर्थं कुरु।

वृत्तमिति: ८-८-८-८।

१. Why CDC
२. Log source
३. Schema
४. Sinks
५. Outbox contrast
६. Bootstrap/lag
७. Privacy
८. Order
९. Deletes
१०. Stream/warehouse
११. Five limbs
१२. Close

1. Prior: outbox 2026-09-28; streams 2026-07-30; warehouse 2026-08-02; idempotency 2026-10-04.
2. Domain: change data capture.
3. English minimizable; Sanskrit-first.
4. Style: no em dash; no Oxford comma in English clauses.